Liver cancer intervention auxiliary system

Abstract

The invention relates to the technical field of medical image processing, in particular to a liver cancer intervention auxiliary system. According to the method, the boundary structure and curvature change of the liver region can be extracted from multi-source images such as MRI, CT and enhanced CT through cooperative processing of multi-modal images, and candidate path segments are constructed step by step by utilizing continuity and direction stability of a boundary turning region; sorting identification is carried out on the candidate path segments in combination with image parameters such as gray distribution, texture features and background contrast, path segment quality levels based on multiple image characteristics are established, and path areas with multi-modal parameter advantages in the image are enhanced and displayed by evaluating response distribution of the sorting levels and carrying out weight adjustment, so that the image quality is improved. And finally, the path information after transparency adjustment or edge enhancement is superposed in the original image frame, so that the image presents a clearer path structure and spatial distribution, and the visual perception capability of a doctor on the direction of the intervention path is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a liver cancer intervention assistance system. Background Art

[0002] The field of medical image processing technology involves the use of computer vision and image analysis methods to process, analyze and identify medical image data to assist doctors in disease diagnosis, treatment planning and postoperative evaluation.

[0003] The traditional liver cancer interventional assistance system is an information system that assists doctors with positioning and path planning during interventional treatments such as transcatheter arterial chemoembolization or radiofrequency ablation for liver cancer patients. It accurately identifies the liver and its internal structures during interventional treatment and assists doctors in locating tumor lesions and vascular orientation.

[0004] Traditional liver cancer interventional assistance systems only identify liver structures based on a single imaging modality and lack in-depth quantification and classification of image boundary features. When faced with images with complex liver shapes or blurred boundaries, they are unable to accurately extract key path turning point information, resulting in problems such as crossing high-risk areas or ignoring tiny lesions during path planning. The path selection process also does not involve refined analysis and sorting control of image parameter levels, so that brightness, texture or background differences in the image cannot be effectively converted into a basis for path judgment. Especially in multimodal image comparison, the system cannot automatically evaluate the relative advantages of the path area under each modality, thereby limiting the visual level of the image enhancement results. Doctors need to rely on subjective judgment when identifying intervention routes, which is easily disturbed by visual perception bias and has limitations in actual surgical path decision-making. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a liver cancer intervention assistance system.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: The liver cancer intervention assistance system includes:

[0007] Curvature grading module: Obtains the contour area of ​​multimodal liver images to identify the external boundary points of the liver, clusters the curvature features of each boundary point, and obtains the liver boundary curvature grading results;

[0008] A path segment construction module is configured to extract a turning curvature region from the liver boundary curvature classification result, and sequentially construct path segments based on continuous boundary point segments of the turning curvature region to obtain a candidate set of interventional path segments;

[0009] Parameter extraction module: obtains the liver region image where the interventional path segment candidate set is located, extracts the corresponding image parameter sequence, sorts the image parameters of the sequence, and obtains the sorted sequence of liver image path segments;

[0010] A weight control module is configured to divide and adjust the weight of each path segment region in the sorted sequence of the liver image path segments to construct a weighted path segment region set;

[0011] Annotation generation module: obtains the spatial coordinate position corresponding to each path segment in the weighted path segment area set, adjusts the transparency or enhances the edge of the path segment area according to the weight order, and superimposes it on the original liver image frame to obtain a liver image intervention auxiliary annotation image.

[0012] As a further solution of the present invention, the liver boundary curvature grading result includes a curvature category mark, a boundary point index, and an area type label; the intervention path segment candidate set includes a path segment starting point, a path segment midpoint, and a path segment end point; the sorting sequence of the liver image path segment includes a grayscale mean sorting position, a texture edge amplitude sorting position, and a background contrast sorting position; the weighted path segment area set includes a path segment sorting structure and a modality corresponding weight; and the liver image intervention auxiliary annotation image includes a path segment spatial position, transparency processing content, and edge display content.

[0013] As a further solution of the present invention, the curvature grading module includes:

[0014] Contour acquisition submodule: Acquires multimodal liver imaging data from MRI, CT, and enhanced CT, detects grayscale edges in the image, identifies the external contours of the liver structure area, extracts liver edge point coordinates in sequential pixel order, and preserves the structural positional relationship to obtain a liver contour boundary point set;

[0015] Curvature extraction submodule: Based on the spatial arrangement relationship between each point in the liver contour boundary point set and its two adjacent points, a vector combination is constructed and the direction change angle is calculated, and the angle value is converted into a boundary point curvature feature sequence;

[0016] Boundary grading submodule: Based on the curvature amplitude and relative change difference of each point in the boundary point curvature feature sequence, it is divided into multiple fluctuation intervals and the boundary layer is marked to generate the liver boundary curvature grading result.

[0017] As a further solution of the present invention, the path segment construction module includes:

[0018] Turning point extraction submodule: calling the boundary point positions marked as turning curvature areas in the liver boundary curvature classification results, extracting adjacent point segments according to the continuous distribution relationship in the boundary point sequence, retaining the point position sequence information of each continuous segment, and obtaining a turning boundary point segment set;

[0019] Path generation submodule: Based on the boundary point sequence contained in the turning boundary point segment set, fixed-length point segments are selected in sequence to construct a path structure, the starting point, midpoint, and end point coordinates of each path segment are extracted, and a structured path segment set is established to obtain a preliminary path segment structure set;

[0020] Continuity screening submodule: Based on the angular relationship between the start and end direction vectors of each segment in the preliminary path segment structure set and the midpoint tangent direction, the direction difference value is calculated and the range distribution within the pass continuity angle threshold is determined. The path segments with stable direction changes are retained to generate a candidate set of intervention path segments.

[0021] As a further solution of the present invention, the parameter extraction module includes:

[0022] The regional positioning submodule calls the spatial coordinates of each path segment in the interventional path segment candidate set, obtains the liver region image at the corresponding position in the MRI, CT and enhanced CT tri-modal images, extracts the image segments covered by each path segment region, and generates a path segment region image set;

[0023] Image feature extraction submodule: Based on the grayscale distribution structure of each area image in the path segment area image set under three modalities, the grayscale mean and background contrast of each area are obtained, the Sobel edge detection algorithm is called to obtain the texture edge amplitude, and the three image features are combined to generate a path segment image parameter sequence;

[0024] Parameter sorting submodule: Based on the values ​​of the three image parameters in the path segment image parameter sequence under MRI, CT and enhanced CT modalities, the sorting position and parameter strength level within the modality are calculated respectively, and each path segment area is sorted and numbered to generate a sorting sequence of liver image path segments.

[0025] As a further solution of the present invention, the weight control module includes:

[0026] The sorting recognition submodule is used to call the sorting position of the grayscale mean, texture edge amplitude, and background contrast of each path segment in the sorting sequence of the liver image path segments under the three modal images, count the sorting level combinations under each modality, and generate path segment sorting position distribution data;

[0027] Contribution evaluation submodule: Based on the ranking hierarchical structure of each path segment in the path segment ranking position distribution data, the path segment regions where all three parameters are in the upper order are identified and marked as high contribution regions, and the path segment regions where two parameters are in the lower order are identified and marked as low contribution regions, and the response levels of the path segments are divided to obtain the path segment discrimination contribution weight distribution data;

[0028] Weight integration submodule: based on the path segment judgment contribution weight distribution data, adjust the participation proportion of the path segment, and generate a weighted path segment area set.

[0029] As a further solution of the present invention, the annotation generation module includes:

[0030] Coordinate extraction submodule: calling the index information of each path segment in the weighted path segment region set, locating the spatial coordinate position in the MRI, CT and enhanced CT trimodal liver images, establishing the correspondence between the path segment region and the image frame, and generating path segment spatial positioning data;

[0031] Image enhancement submodule: according to the spatial positioning data of the path segments and the corresponding weight sorting order, select transparency adjustment or edge enhancement to process each weight path segment respectively, and obtain the path segment area enhancement result;

[0032] Image fusion submodule: Based on the image segments of all path segments in the path segment area enhancement results, the image segments are sequentially superimposed on the original trimodal liver image frame according to the coordinate positions of the image segments, and the path structure characteristics, boundary performance and multimodal parameter differences are fused to generate a liver image intervention auxiliary annotation image.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are:

[0034] In the present invention, through the collaborative processing of multimodal images, the boundary structure and curvature changes of the liver area can be extracted in multi-source images such as MRI, CT and enhanced CT, and the continuity and directional stability of the boundary turning area are used to gradually construct candidate path segments. The candidate path segments are sorted and identified in combination with image parameters such as grayscale distribution, texture characteristics and background contrast, and a path segment quality ranking based on multiple image characteristics is established. By evaluating the response distribution of the sorting hierarchy and adjusting the weights, the path area with multimodal parameter advantages in the image is enhanced. Finally, the path information after transparency adjustment or edge enhancement is superimposed on the original image frame, so that the image presents a clearer path structure and spatial distribution, thereby improving the doctor's visual perception of the direction of the interventional path. While improving the accuracy of path planning, it effectively avoids the influence of low-contribution areas on the image interference, ensuring the reliability and practicality of the tumor and vascular pathway identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a system flow chart of the present invention;

[0036] Figure 2 This is a flow chart of the curvature grading module of the present invention;

[0037] Figure 3A flow chart of a path segment construction module of the present invention;

[0038] Figure 4 This is a flow chart of the parameter extraction module of the present invention;

[0039] Figure 5 This is a flow chart of the weight control module of the present invention;

[0040] Figure 6 The flowchart of the label generation module of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0042] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0043] See also Figure 1 , liver cancer interventional assistance system includes:

[0044] Curvature grading module: Obtains the contour area of ​​multimodal liver images to identify the external boundary points of the liver, clusters the curvature features of each boundary point, and obtains the liver boundary curvature grading results;

[0045] Pathway segment construction module: Extracts the turning curvature region from the liver boundary curvature grading results, and constructs path segments in sequence based on the continuous boundary point segments of the turning curvature region to obtain a candidate set of interventional path segments;

[0046] Parameter extraction module: obtains the liver region image where the interventional path segment candidate set is located, extracts the corresponding image parameter sequence, sorts the image parameters of the sequence, and obtains the sorted sequence of liver image path segments;

[0047] Weight control module: divides and adjusts the weight of each path segment area in the sorted sequence of liver image path segments to construct a weighted path segment area set;

[0048] Annotation generation module: Obtain the spatial coordinate position corresponding to each path segment in the weighted path segment area set, adjust the transparency or enhance the edge of the path segment area according to the weight order, and superimpose it on the original liver image frame to obtain the liver image intervention auxiliary annotation image;

[0049] The liver boundary curvature grading results include curvature category labels, boundary point indexes, and region type labels. The interventional path segment candidate set includes the path segment starting point, path segment midpoint, and path segment end point. The sorting sequence of liver image path segments includes the grayscale mean sorting position, texture edge amplitude sorting position, and background contrast sorting position. The weighted path segment region set includes the path segment sorting structure and modality corresponding weights. The liver image intervention auxiliary annotation image includes the path segment spatial position, transparency processing content, and edge display content.

[0050] See also Figure 2 , the curvature grading module includes:

[0051] Contour acquisition submodule: Acquires multimodal liver imaging data from MRI, CT, and enhanced CT, detects grayscale edges in the image, identifies the external contours of the liver structure area, extracts liver edge point coordinates in sequential pixel order, and preserves the structural positional relationship to obtain a liver contour boundary point set;

[0052] Multimodal liver imaging data from MRI, CT, and enhanced CT were acquired. The three-modal image sequences were read separately during the image loading process and uniformly registered. All modalities were aligned in the same two-dimensional plane coordinates with a unified sampling resolution of 1 mm pixel pitch. The image frame range was restricted to the sequence segment containing the liver anatomical region. A detection threshold for the region with increasing edge grayscale gradient was set to extract the pixel sideband region with significant grayscale changes in the image. After determining whether the region formed a closed connected image structure, an edge tracking operation was called to extract the pixel sequence of the closed region sideband. The coordinate values ​​of the pixels were recorded according to their row and column order in the image frame, and a continuous sequence mapping structure was established according to the curve direction. The boundary connectivity of the extracted boundary points was judged, and isolated noise segments were deleted. Only the main edge path in the closed boundary region was retained. The continuous outer edge pixel coordinates extracted from the overlapping region after uniform registration in the three-modal images were summarized and collated to obtain the liver contour boundary point set.

[0053] Curvature extraction submodule: Based on the spatial arrangement relationship between each point in the liver contour boundary point set and its two adjacent points, it constructs a vector combination and calculates the angle of direction change, converting the angle value into a boundary point curvature feature sequence;

[0054] Based on the spatial arrangement relationship between each point in the liver contour boundary point set and its adjacent points before and after, two two-dimensional direction vectors pointing from the current point to the front point and the back point are constructed in sequence, and the degree of direction change of each boundary point is characterized by calculating the cosine value of the angle between the two vectors. Assume that the current boundary point is , its previous neighbor is , the next neighbor is , then define:

[0055] The first vector (from the previous point to the current point) is ;

[0056] The second vector (from the current point to the next point) is .

[0057] The cosine formula of the angle between two vectors is expressed as:

[0058] ;

[0059] in, : The horizontal coordinate of the first point (previous point) in the two-dimensional plane of the image; : The vertical coordinate of the first point (previous neighbor) in the two-dimensional plane of the image; : The horizontal coordinate of the current boundary point; : The vertical coordinate of the current boundary point; : The horizontal coordinate of the second point (the next adjacent point); : The vertical coordinate of the second point (the next adjacent point); :from arrive The direction vector represents the boundary direction from the previous neighbor to the current point; :from arrive The direction vector represents the boundary direction from the current point to the next neighboring point; : The dot product of two direction vectors is used to calculate the angle between them; : The modulus of the first vector, i.e. arrive The Euclidean distance of : The modulus of the second vector, i.e. arrive The Euclidean distance of : The angle between the two vectors, indicating the degree of local direction change of the boundary; : The cosine value of the angle. The closer the value is to 1, the smoother the boundary is, and the smaller the value is, the more obvious the turning point is.

[0060] Suppose the boundary point sequence is 、 、 ,but: ; ; Dot product: ; Mould length: ; ; Cosine value calculation: , indicating that the boundary direction at that point is continuous, without obvious transitions. During boundary curvature extraction, the cosine value is used to determine the degree of directional change at a boundary point. Its value ranges from 0 to 1, representing the angle between adjacent boundary direction vectors. The boundary direction changes corresponding to different values ​​are as follows: when the cosine value is close to 1, for example, above 0.95, it indicates that the two direction vectors are almost in the same direction, that is, the boundary segments are arranged in an approximately straight line. The direction continuity at the current boundary point is good, the boundary is relatively smooth, and there is no obvious turning point. When the cosine value is in the medium range, for example, between 0.7 and 0.95, it indicates that there is a certain angle change between adjacent boundary directions. In this case, the boundary may have a relatively gentle turning point, which is often seen at the physiological bends of organ contours and belongs to the structural direction adjustment area. When the cosine value is lower than 0.7, it indicates that the angle between the two direction vectors is large and there is a significant directional deviation. The current boundary point may be a structural mutation location such as a sharp corner, protrusion or depression, which belongs to the boundary discontinuity or high variation area. If the cosine value is close to 0, it means that the direction vectors are approximately perpendicular, indicating that the boundary has a sharp angle change of nearly 90 degrees. Such points are rare in anatomical structures and usually represent image abnormalities or errors in boundary detection. They need to be screened out or corrected in combination with subsequent processing. Therefore, the degree of directional change at a boundary point can be directly determined based on the magnitude of the cosine value, and the boundary region can be divided into a smooth segment, a turning segment, or a sudden segment accordingly. This judgment logic is universally applicable to image contour processing and is particularly suitable for the liver, an organ with complex structure and irregular boundaries in medical images.

[0061] Repeat the above operation for all boundary points to form a direction change sequence for the entire boundary. Mapping to curvature index ,definition: ,in: : No. The curvature value of each boundary point reflects the degree of direction change; : The cosine value of the direction angle of the point, the value range For example, when When , indicating that there is a certain angle mutation at this point. Summarize into a continuous numerical sequence to form the boundary point curvature feature sequence.

[0062] Boundary grading submodule: Based on the curvature amplitude and relative change difference of each point in the boundary point curvature feature sequence, it is divided into multiple fluctuation intervals and the boundary layer is marked to generate the liver boundary curvature grading results;

[0063] According to the change of curvature value of each boundary point in the boundary point curvature feature sequence, the curvature sequence was first traversed in point order to detect its local fluctuation amplitude. Every five consecutive points were used as sliding windows, and the difference between the maximum and minimum values ​​in the window was calculated to reflect the curvature fluctuation range of the local area. The fluctuation values ​​of all windows were then recorded and the global average fluctuation level was calculated. A preliminary judgment was made based on the ratio of the fluctuation value in each window to the global average. When the window fluctuation value was greater than 1.5 times the global mean, it was classified as a mutation area; when it was less than 0.5 times, it was classified as a gentle area. The remaining segments were marked as turning areas. At the same time, a connectivity test was performed on each type of segment to ensure that the curvature level in the same segment was consistent and continuously distributed. Finally, a curvature level label was established for each point in the boundary point sequence, marking them as gentle boundary points, turning boundary points, and mutation boundary points respectively. The result table was output with the point sequence as the index to generate the liver boundary curvature grading result.

[0064] See also Figure 3 , the path segment building blocks include:

[0065] Turning point extraction submodule: Call the boundary point positions marked as turning curvature areas in the liver boundary curvature classification results, extract adjacent point segments according to the continuous distribution relationship in the boundary point sequence, retain the point position sequence information of each continuous segment, and obtain the turning boundary point segment set;

[0066] Call the boundary point positions marked as turning curvature areas in the liver boundary curvature classification results, traverse them in the order of the point numbers in the boundary sequence, first perform a difference operation on the point index to identify whether there is number continuity between adjacent points, for example, the point numbers are 105, 106, 107, 109 , it can be judged that there is an interruption at number 109, so 105 to 107 can be regarded as a continuous segment. If the subsequent numbers continue to increase, such as 110, 111, and 112, a new segment can be opened and accumulated from 109. The judgment basis is whether the difference between the adjacent point numbers is 1. If it is equal to 1, they belong to the same segment, otherwise it is a segment interruption. At the same time, the minimum threshold of the segment length is set to 5 to avoid misjudgment of path construction due to short segment data. For boundary segments that meet continuity and are not less than 5 points in length, their starting point number, end point number and all coordinate sequence information are recorded one by one. The coordinate format retains the two-dimensional image plane position. Then, the connectivity check is performed on the extracted multiple segments, and the segments containing isolated points or non-turning labels are eliminated. The number index and point coordinate data of all qualified segments are summarized to form a turning boundary point segment set.

[0067] Path generation submodule: Based on the boundary point sequence contained in the turning boundary point segment set, fixed-length point segments are selected in sequence to construct the path structure, the starting point, midpoint and end point coordinates of each path segment are extracted, and a structured path segment set is established to obtain a preliminary path segment structure set;

[0068] Based on the point coordinate sequence in each turning segment in the turning boundary point segment set, the candidate path segment structure is constructed according to a fixed length. Each path segment is set to contain 7 points. When traversing each segment, sliding window processing is used, and the step size is set to 3. That is, starting from the 1st point in the same segment, the 1st to 7th points are extracted to form path segment 1, and then a group of points are extracted from the 4th point to form path segment 2. Slide in sequence until there are less than 7 points remaining. For example, if there are point sequence numbers 201 to 212 in a turning segment, the path segments can be constructed in sequence: 201-207, 204-210, 207-213, where the starting point is the 1st point, the midpoint is the 4th point, and the end point is the 7th point. The corresponding coordinates are as follows: , the midpoint is , the end point is , the three-point structure of the path segment can be recorded as: starting point coordinates = 32, 45; midpoint coordinates = 35, 48; end point coordinates = 38, 50, and a unique ID value such as Path_001 is assigned to the path segment. At the same time, the turning area number to which the segment belongs is recorded for subsequent tracking. All path segments are organized into a structured list in the order of numbers, including fields such as the three-point coordinates, the boundary segment number to which they belong, and the sequence index. All path segment information is constructed and summarized to obtain a preliminary path segment structure set.

[0069] Continuity screening submodule: Based on the angle relationship between the start and end direction vectors and the midpoint tangent direction of each segment in the preliminary path segment structure set, the direction difference value is calculated and the range distribution within the continuity angle threshold is determined. Path segments with stable direction changes are retained to generate a candidate set of intervention path segments;

[0070] Based on the three-point coordinate structure of each path segment in the preliminary path segment structure set, the starting point, midpoint and end point of the path segment are called respectively. According to the vector calculation result, the cosine value of the angle between the starting and ending direction vectors of the path segment and the tangent direction vector of the midpoint is obtained. The direction difference value is calculated using the formula:

[0071] ;

[0072] in: : represents the cosine value of the starting point direction vector and the midpoint tangent direction; : represents the cosine value of the midpoint tangent direction and the end point direction vector; : The direction difference value of the current path segment, which measures the directional consistency of the segments on both sides of the path. The closer the value is to 0, the smoother the direction change.

[0073] For example, suppose the coordinates of the starting point, midpoint, and end point of a path segment are 、 、 The cosine of the angle between the vector from the starting point to the midpoint and the tangent direction of the midpoint is The cosine of the angle between the midpoint tangent and the end point direction is , the direction difference value is calculated as follows: , compare the direction difference value with the set passage continuity angle threshold, set the threshold to 0.15, and the reference basis is that in the image structure path, if the difference in the cosine value of the angle between the starting and ending directions of the path and the midpoint direction is less than 0.15, the direction tends to be consistent and meets the continuous passage condition. This value is derived from the statistics of 100 real path segment samples, of which more than 85% of the direction cosine difference values ​​of the continuous segments on the structure are between [0.05, 0.13]. Therefore, 0.15 is set as the upper limit for judging passage stability. The judgment conditions are as follows: If , then the path segment direction is stable and retained; if , the path segment direction changes too much and is removed. This process is continued for all path segments in the preliminary path segment structure set. The directional difference values ​​of each path segment are calculated and threshold judgment is performed. All path segments that meet the directional stability conditions are uniformly numbered and output. The set data of the retained segments is constructed to generate the candidate set of intervention path segments.

[0074] See also Figure 4 , the parameter extraction module includes:

[0075] The regional positioning submodule calls the spatial coordinates of each path segment in the interventional path segment candidate set, obtains the liver region image at the corresponding position in the MRI, CT, and enhanced CT tri-modality images, extracts the image fragments covered by each path segment area, and generates a path segment area image set;

[0076] The three-point coordinate information of each path segment in the intervention path segment candidate set was called, and the image area corresponding to the coordinate position was retrieved in the MRI, CT and enhanced CT three-modality liver image frames. The size range of the path segment area interception box was set to extend 5 pixels to the left and right of the center point and 5 pixels to the top and bottom to form a A regional window is used to extract the image grayscale matrix within the window range under each modality to ensure that the image capture area contains the complete path segment structure area. The pixel offset between the modalities is aligned and the coordinate mapping table is used to unify the regional index structure. After extraction, the image content is retained as a standard format matrix array. Each path segment corresponds to a regional image under each modality, forming a total of three image fragments. The image content of all path segments under each modality is summarized and an index relationship is established to form a path segment regional image set.

[0077] Image feature extraction submodule: Based on the grayscale distribution structure of each region image in the path segment region image set under three modalities, the grayscale mean and background contrast of each region are obtained, the Sobel edge detection algorithm is called to obtain the texture edge amplitude, and the three image features are combined to generate a path segment image parameter sequence;

[0078] Based on the grayscale distribution of each regional image in the path segment region image set, feature calculation is performed on each regional image under MRI, CT and enhanced CT modalities in turn. First, the total image pixel value is divided by the number of pixels to calculate the grayscale mean of the region. Then, background pixel blocks are extracted around the image region, their grayscale mean is calculated and the difference is taken to obtain the background contrast. Then, the Sobel edge detection method is called to process the original image region, and its horizontal and vertical gradient maps are extracted and the gradient amplitude of the corresponding pixel position is calculated. The average gradient amplitude of all pixels is taken as the background contrast. is the regional texture edge amplitude. The three features including grayscale mean, background contrast and texture amplitude respectively form a feature vector. For example, the regional grayscale mean of a path segment under CT modality is 138.2, and the background grayscale is 112.7, then the contrast is 25.5, and the texture edge amplitude is 18.6. The image feature vector of this path segment under CT modality is [138.2, 25.5, 18.6]. Repeat the feature extraction process for all path segments under all modalities, unify the structured records, and summarize them into a path segment image parameter sequence.

[0079] Parameter sorting submodule: Based on the values ​​of the three image parameters in the path segment image parameter sequence under MRI, CT, and enhanced CT modalities, the sorting position and parameter strength level within the modality are calculated respectively, and each path segment area is sorted and numbered to generate a sorting sequence of liver image path segments;

[0080] According to the three image features of grayscale mean, background contrast and texture edge amplitude extracted from each path segment in the path segment image parameter sequence under MRI, CT and enhanced CT, all path segments with the same parameter dimension in each modality are numerically compared and sorted. The sorting positions are numbered in ascending order. The larger the value, the higher the sorting position, forming a sorting sequence from 1 to N, where N is the total number of path segments. The sorting position of each path segment in each parameter dimension is recorded as ,in: : path segment number; : Indicates parameter categories, corresponding to grayscale mean, background contrast and texture edge amplitude respectively; : Indicates image modality. After the ranking position is determined, the parameters are further divided into strong and weak levels, and the ranking level interval is set as: the top 25% ranking position (i.e. ) is determined as the upper level; the middle 50% ranking position (i.e. ) is determined to be the middle level; the last 25% ranking position (i.e. ) is judged as the lower level.

[0081] If the total number of path segments , the sorting level cutoff points are as follows: upper-level interval: sorting position 1–5; middle-level interval: sorting position 6–15; lower-level interval: sorting position 16–20.

[0082] Assume that for the seventh path segment in CT mode, the three parameters are: grayscale mean 141.3, ranking 3rd among all 20 path segments; background contrast 17.4, ranking 12th; texture edge amplitude 11.2, ranking 18th. The ranking level of this path segment is: grayscale mean: high (3∈1–5); background contrast: medium (12∈6–15); texture amplitude: low (18∈16–20).

[0083] The three-item ranking hierarchy was recorded as [upper order, middle order, lower order], and each path segment was assigned a three-item ranking position and a strength and weakness level identifier in each modality. The ranking results for all path segments across all modalities were numbered and output to generate a ranking sequence for liver imaging path segments.

[0084] See also Figure 5 , the weight control module includes:

[0085] The sorting and recognition submodule calls the sorting sequence of liver image path segments based on the grayscale mean, texture edge amplitude, and background contrast of each path segment in the three modal images, calculates the sorting level combination in each modality, and generates path segment sorting position distribution data;

[0086] The sorting results of the three image parameters of each path segment in the sorting sequence of liver image path segments under the three modalities of MRI, CT, and enhanced CT were called, and the sorting positions of their grayscale mean, texture edge amplitude, and background contrast were read in turn. The sorting position level corresponding to each parameter in each modality was recorded as upper, middle, or lower. The path segments were grouped and counted according to the path segment ID to form a set of hierarchical combinations corresponding to the three parameters in each modality. For example, if the sorting level of the three parameters of a path segment is [upper, middle, lower] in the MRI modality, [middle, upper, upper] in the CT modality, and [middle, middle, lower] in the enhanced CT modality, then the path segment contains a total of 9 sorting hierarchical information under the three modalities. This information is structured and stored, and the sorting level of each parameter is marked. A unified path segment sorting information table is established, and the path segment sorting position distribution data is generated using the path segment as the index.

[0087] Contribution evaluation submodule: Based on the ranking hierarchical structure of each path segment in the path segment ranking position distribution data, the path segment areas where all three parameters are in the upper position are identified and marked as high contribution areas, and the path segment areas where two parameters are in the lower position are identified and marked as low contribution areas. The response levels of the path segments are divided to obtain the path segment discrimination contribution weight distribution data;

[0088] According to the nine ranking level combinations of each path segment under three modes in the path segment ranking position distribution data, the number of "upper order" and "lower order" ranking levels in each path segment are counted. When the number of upper orders in the nine ranking levels of a path segment is equal to 9, it is judged to be an extremely high order; when all three parameters are upper orders under any mode, it is judged to be a high contribution area; if two of the ranking levels of a path segment are lower orders and it does not contain any upper orders, it is judged to be a low contribution area; the rest are classified as general areas. The grade classification of the path segment is executed in sequence according to the above logic and marked as "high", "medium" and "low". The grade label is recorded for each path segment and its belonging classification is output to generate the path segment discrimination contribution weight distribution data.

[0089] Weight integration submodule: Based on the path segment, the level of each path segment in the contribution weight distribution data is determined, the participation ratio of the path segment is adjusted, and a weighted path segment area set is generated;

[0090] Based on the response level label of each path segment in the path segment discrimination contribution weight distribution data, the comprehensive participation proportion of the path segment is adjusted to calculate the comprehensive weight value of the path segment. , using ranking score and grade control factors The product relationship is constructed, and the formula is defined as: ,in, : No. The comprehensive weighted value of each path segment is used to determine its importance in the weighted path segment area set; : No. The level control factor of each path segment is set according to whether its ranking structure meets the conditions of high contribution or low contribution; : No. The image parameter ranking score of each path segment is used to comprehensively reflect the advantage of the path segment in the image parameter ranking of each modality.

[0091] Ranking score The calculation formula is as follows:

[0092] ;

[0093] The meaning of each parameter is: :Indicates the path segments; : Image parameter category number, value range ,For example , corresponding to three types of parameters: image grayscale mean, texture edge amplitude, and background contrast; : Mode number, value range ,For example , representing the three imaging modalities of MRI, CT and enhanced CT respectively; : total number of path segments; : No. The path segment in Parameters, The ranking position under each mode (1 is the best and N is the worst); : It is the reverse normalized value of the ranking score, which is used to increase the influence weight of the top-ranked path segments.

[0094] Level control factor The setting rule is: if the path segment has three image parameters in the upper level (i.e., the ranking position is in the top 25% of the mode) in any mode, it is considered a high contribution area. ,in If any two image parameters of a path segment are at the lower level (i.e., the ranking position is in the last 25% of the modalities) in all modalities and there is no upper-order item, it is considered a low-contribution area. ,in ; Other path segments are considered as medium response areas, set ,in .

[0095] Set the total number of path segments , No. The image parameters of the path segments are sorted as follows ( : Grayscale mean, : Texture edge amplitude, : background contrast; :MRI, :CT, : Enhanced CT): , , (MRI), , , (CT), , , (Enhanced CT).

[0096] Its ranking score is: Since the three parameters of this path segment are in the top 5, that is, the top 25% interval, under the MRI modality, it is marked as a high contribution area and assigned , the comprehensive weight is: , repeat this calculation process and calculate the weight of all path segments one by one and with the set minimum retention threshold For comparison, it is recommended to set For all 70% of the average value, that is, if the average value is 6.0, then , only keep The path segments enter the output set to form a weighted path segment area set.

[0097] See also Figure 6 , the annotation generation module includes:

[0098] Coordinate extraction submodule: calls the index information of each path segment in the weighted path segment region set, locates the spatial coordinate position in the MRI, CT, and enhanced CT liver images, establishes the correspondence between the path segment region and the image frame, and generates the path segment spatial positioning data;

[0099] The unique index number of each path segment in the weighted path segment area set is called to extract the corresponding path structure coordinate data, including the starting point, midpoint and end point positions of the path segment. Through the preset three-modal image frame registration relationship, each coordinate point is mapped to the image frame number and pixel position in MRI, CT and enhanced CT images respectively. The affine transformation parameters are used to unify the coordinates between the modalities, and a mapping table between path segments and image frames is constructed. The image frame index corresponding to each path segment under the three modalities and its relative position offset are recorded to ensure the consistency of coordinate matching. At the same time, the original weight value and sorting number of each path segment are retained as additional attribute fields. A structured correspondence table of fields such as path segment number, modality type, frame position, coordinate range, and weight level is established to generate path segment spatial positioning data.

[0100] Image enhancement submodule: Based on the spatial positioning data of the path segments and the corresponding weight sorting order, the transparency adjustment or edge enhancement method is selected to process each weight path segment separately to obtain the path segment area enhancement result;

[0101] According to the coordinate range and weight ranking information of each path segment in the path segment spatial positioning data on the trimodal image, the image enhancement methods corresponding to different weight levels are set. Among them, the path segments ranked in the top 25% of the weight are subjected to edge enhancement processing, and the Sobel operator is called to extract the boundary area and enhance the grayscale contrast to strengthen the contour performance; for the path segments with weights in the middle 50%, the original image is kept unchanged and directly marked; for the path segments with weights in the bottom 25%, transparency adjustment is performed, and their grayscale values ​​are reduced by 20% overall, and a semi-transparent mask is superimposed to ensure that they exist in a faded form in the image frame. The enhanced image fragments retain the path segment number and modal position label. The three processing methods uniformly generate a set of processed image blocks, and output the image structure data of all path segment enhancement results to form the path segment area enhancement result.

[0102] Image fusion submodule: Based on the image segments of all path segments in the path segment region enhancement results, the image segments are sequentially superimposed onto the original three-modal liver image frame according to their coordinate positions. The path structure features, boundary expressions, and multimodal parameter differences are integrated to generate an intervention-assisted annotation image of the liver image.

[0103] Based on the image enhancement fragments of each path segment in the path segment area enhancement results and their corresponding spatial positioning information, they are grouped by modality type and sequentially superimposed on the original liver image frames of MRI, CT, and enhanced CT. If there are multiple path segments covering the same area, image synthesis is performed according to their weight ranking priority. High-weight path segments are rendered first to retain their enhancement effects. Pixel-level synthesis is performed in the presence of transparent path segment areas to ensure continuity and coverage consistency between different processing levels. During the fusion process, the path segment structural information and modality mapping position are simultaneously retained. The image frames after fusion of each modality are numbered, stored, and output to generate liver imaging intervention-assisted annotation images that integrate structural features, enhancement performance, and modality difference information.

[0104] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. Liver cancer interventional assistance system, characterized by: The system comprises: Curvature grading module: Obtains the contour area of ​​multimodal liver images to identify the external boundary points of the liver, clusters the curvature features of each boundary point, and obtains the liver boundary curvature grading results; A path segment construction module is configured to extract a turning curvature region from the liver boundary curvature classification result, and sequentially construct path segments based on continuous boundary point segments of the turning curvature region to obtain a candidate set of interventional path segments; Parameter extraction module: obtains the liver region image where the interventional path segment candidate set is located, extracts the corresponding image parameter sequence, sorts the image parameters of the sequence, and obtains the sorted sequence of liver image path segments; A weight control module is configured to divide and adjust the weight of each path segment region in the sorted sequence of the liver image path segments to construct a weighted path segment region set; Annotation generation module: obtains the spatial coordinate position corresponding to each path segment in the weighted path segment area set, adjusts the transparency or enhances the edge of the path segment area according to the weight order, and superimposes it on the original liver image frame to obtain a liver image intervention auxiliary annotation image.

2. The liver cancer intervention assistance system according to claim 1, characterized in that: The liver boundary curvature grading result includes a curvature category mark, a boundary point index, and a region type label; the intervention path segment candidate set includes a path segment starting point, a path segment midpoint, and a path segment end point; the sorting sequence of the liver image path segment includes a grayscale mean sorting position, a texture edge amplitude sorting position, and a background contrast sorting position; the weighted path segment region set includes a path segment sorting structure and a modality corresponding weight; and the liver image intervention auxiliary annotation image includes a path segment spatial position, transparency processing content, and edge display content.

3. The liver cancer intervention assistance system according to claim 1, characterized in that: The curvature grading module includes: Contour acquisition submodule: Acquires multimodal liver imaging data from MRI, CT, and enhanced CT, detects grayscale edges in the image, identifies the external contours of the liver structure area, extracts liver edge point coordinates in sequential pixel order, and preserves the structural positional relationship to obtain a liver contour boundary point set; Curvature extraction submodule: Based on the spatial arrangement relationship between each point in the liver contour boundary point set and its two adjacent points, a vector combination is constructed and the direction change angle is calculated, and the angle value is converted into a boundary point curvature feature sequence; Boundary grading submodule: Based on the curvature amplitude and relative change difference of each point in the boundary point curvature feature sequence, it is divided into multiple fluctuation intervals and the boundary layer is marked to generate the liver boundary curvature grading result.

4. The liver cancer intervention assistance system according to claim 3, characterized in that: The path segment construction module includes: Turning point extraction submodule: calling the boundary point positions marked as turning curvature areas in the liver boundary curvature classification results, extracting adjacent point segments according to the continuous distribution relationship in the boundary point sequence, retaining the point position sequence information of each continuous segment, and obtaining a turning boundary point segment set; Path generation submodule: Based on the boundary point sequence contained in the turning boundary point segment set, fixed-length point segments are selected in sequence to construct a path structure, the starting point, midpoint, and end point coordinates of each path segment are extracted, and a structured path segment set is established to obtain a preliminary path segment structure set; Continuity screening submodule: Based on the angular relationship between the start and end direction vectors of each segment in the preliminary path segment structure set and the midpoint tangent direction, the direction difference value is calculated and the range distribution within the pass continuity angle threshold is determined. The path segments with stable direction changes are retained to generate a candidate set of intervention path segments.

5. The liver cancer intervention assistance system according to claim 4, characterized in that: The parameter extraction module includes: The regional positioning submodule calls the spatial coordinates of each path segment in the interventional path segment candidate set, obtains the liver region image at the corresponding position in the MRI, CT and enhanced CT tri-modal images, extracts the image segments covered by each path segment region, and generates a path segment region image set; Image feature extraction submodule: Based on the grayscale distribution structure of each area image in the path segment area image set under three modalities, the grayscale mean and background contrast of each area are obtained, the Sobel edge detection algorithm is called to obtain the texture edge amplitude, and the three image features are combined to generate a path segment image parameter sequence; Parameter sorting submodule: Based on the values ​​of the three image parameters in the path segment image parameter sequence under MRI, CT and enhanced CT modalities, the sorting position and parameter strength level within the modality are calculated respectively, and each path segment area is sorted and numbered to generate a sorting sequence of liver image path segments.

6. The liver cancer intervention assistance system according to claim 5, characterized in that: The weight control module includes: The sorting recognition submodule is used to call the sorting position of the grayscale mean, texture edge amplitude, and background contrast of each path segment in the sorting sequence of the liver image path segments under the three modal images, count the sorting level combinations under each modality, and generate path segment sorting position distribution data; Contribution evaluation submodule: Based on the ranking hierarchical structure of each path segment in the path segment ranking position distribution data, the path segment regions where all three parameters are in the upper order are identified and marked as high contribution regions, and the path segment regions where two parameters are in the lower order are identified and marked as low contribution regions, and the response levels of the path segments are divided to obtain the path segment discrimination contribution weight distribution data; Weight integration submodule: based on the path segment judgment contribution weight distribution data, adjust the participation proportion of the path segment, and generate a weighted path segment area set.

7. The liver cancer intervention assistance system according to claim 6, characterized in that: The annotation generation module includes: Coordinate extraction submodule: calls the index information of each path segment in the weighted path segment region set, locates the spatial coordinate position in the MRI, CT and enhanced CT three-modality liver images, establishes the correspondence between the path segment region and the image frame, and generates path segment spatial positioning data; Image enhancement submodule: according to the spatial positioning data of the path segments and the corresponding weight sorting order, select transparency adjustment or edge enhancement to process each weight path segment respectively, and obtain the path segment area enhancement result; Image fusion submodule: Based on the image segments of all path segments in the path segment area enhancement results, the image segments are sequentially superimposed on the original trimodal liver image frame according to the coordinate positions of the image segments, and the path structure characteristics, boundary performance and multimodal parameter differences are fused to generate a liver image intervention auxiliary annotation image.

Images