An adjacent area delineation assistance method before hepatic segmentectomy

Abstract

The present invention relates to the field of image feature recognition technology, and in particular to a method for assisting in outlining adjacent areas before liver segment resection. The method analyzes the CT images at each level of the liver, and quantifies the degree of respiratory influence using the shape change and area change of the vascular area. By performing matching analysis on the target vascular areas between adjacent levels, the changing trend characteristics of the respiratory influence are determined, and then all target vascular areas are traversed by matching and then recombining. Based on the changing trend characteristics, the optimal combination method representing the portal vein can be determined, and the vascular area represented by the portal vein can be screened out, and effective assistance in outlining adjacent areas can be achieved through three-dimensional reconstruction. The present invention can effectively assist in outlining adjacent areas by analyzing the correlation of vascular areas in CT images of different levels, screening out accurate portal vein areas, and performing three-dimensional reconstruction.

Description

Technical Field

[0001] The present invention relates to the technical field of image feature recognition, and in particular to an auxiliary method for delineating adjacent areas before liver segment resection. Background Art

[0002] Delineation of adjacent regions before liver segment resection is a core component of surgical planning. Its technical development has evolved from traditional two-dimensional image analysis to three-dimensional visualization and intelligent segmentation. Currently, delineation of adjacent regions before liver segment resection can be combined with comprehensive imaging techniques, intelligent algorithms, and intraoperative dynamic navigation. Existing technologies utilize static liver CT image segmentation to remove the influence of adjacent organs such as the spine and ribs, observe the relationship between the tumor and surrounding blood vessels, plan the resection range, and improve segmentation accuracy. However, each liver segment has its own vascular inflow, outflow, and bile drainage. The center of each liver segment has branches of the portal vein, hepatic artery, and bile duct. Vascular outflow from each liver segment is also present through the hepatic vein. Therefore, the hepatic vein and portal vein of different liver segments are susceptible to varying degrees of deformation due to respiratory motion and intraoperative traction, blurring the boundaries of adjacent regions. Failure to effectively distinguish and segment the vascular regions on CT images can lead to inaccurate delineation results. Summary of the Invention

[0003] In order to solve the technical problem that the existing technology cannot effectively segment portal vein information based on liver CT images, resulting in inaccurate delineation of adjacent areas, the purpose of the present invention is to provide an auxiliary method for delineating adjacent areas before liver segmentectomy. The technical solution adopted is as follows:

[0004] The present invention proposes an auxiliary method for delineating adjacent areas before liver segment resection, the method comprising:

[0005] Obtaining multiple consecutive CT images of the patient's liver at each layer of the CT scan, wherein the vascular region in the first frame of the CT image is used as the target vascular region;

[0006] Obtain the degree of spindle-shaped features of each vascular region; at each layer, the vascular regions in adjacent CT frames are matched sequentially, and the degree of respiratory influence of the target vascular region is obtained based on the area change and spindle-shaped feature change between the matched regions;

[0007] Matching target vascular regions of the same liver segment between adjacent slices, forming a set of matching target vascular regions with matching relationships; obtaining a mutual extension probability of the set of matching target vascular regions based on the distance and position distribution between the regions;

[0008] For each liver segment, all matching target vascular region sets contained therein are traversed and combined, and arranged according to mutual extension probability to obtain multiple combinations. The changing trend characteristics of the respiratory influence degree of the matching target vascular region set in each combination are obtained. The optimal combination is screened based on the changing trend characteristics, the number of layers contained in the matching target vascular region set, the degree of fusiform characteristics of the target vascular region corresponding to the matching target vascular region set, and the number of elements in the combination.

[0009] The target vascular area corresponding to the optimal combination is used as the portal vein branch and three-dimensionally reconstructed to assist in delineating the adjacent area.

[0010] Furthermore, the method for obtaining the degree of the spindle-shaped feature includes:

[0011] The minimum circumscribed rectangle of the blood vessel region is obtained, the aspect ratio of the minimum circumscribed rectangle is used as the initial spindle feature degree, and the product of the average gray value of the blood vessel region and the initial spindle feature degree is used as the spindle feature degree.

[0012] Furthermore, the blood vessel regions in adjacent CT image frames are matched in sequence, including:

[0013] For each pair of adjacent frames at each layer, the degree of overlap between the matched vascular region in the previous CT image and the vascular region in the next CT image is obtained, and the vascular region with the highest overlap is selected as the new matched vascular region; the matched vascular region in the first CT image is any target vascular region;

[0014] Match target vascular regions that share the same liver segment between two slices, including:

[0015] For any two adjacent layers, the first frame CT images between the two layers are matched, and a group of target vascular regions with the largest degree of overlap are selected as matching target vascular region pairs with a matching relationship; all layers are traversed, and the target vascular regions contained in the matching target vascular region pairs with a matching relationship are combined into a matching target vascular region set.

[0016] Furthermore, the method for obtaining the degree of respiratory influence includes:

[0017] For each group of adjacent frames, the difference in the average degree of spindle-shaped features and the difference in average area between the matched vascular regions in the previous frame and the next frame are obtained to obtain the morphological distortion degree; the morphological distortion degrees of all groups of adjacent frames are counted, and the degree of respiratory influence is obtained based on the average morphological distortion degree and the morphological distortion degree variance.

[0018] Furthermore, the method for obtaining the morphological distortion degree includes:

[0019] For each group of adjacent frames, a first ratio of the average degree of spindle-shaped features between the matched blood vessel region in the previous frame and the matched blood vessel region in the next frame, and a second ratio of the average areas are obtained; a first product between the first ratio and the second ratio is obtained, and the difference between the first product and the positive integer 1 is normalized to obtain the morphological distortion degree.

[0020] Furthermore, the method for obtaining the mutual extension probability includes:

[0021] In the set of matching target vascular regions, for two matching target vascular regions corresponding to adjacent slices, obtaining the minimum circumscribed rectangles of the two matching target vascular regions, extending the sides of the two minimum circumscribed rectangles, and using the minimum angle formed by the extended lines as the posture difference angle; obtaining a distance between the two matching target vascular regions, calculating a Euclidean norm based on the distance and the posture difference angle, performing negative correlation mapping on the Euclidean norm, and obtaining an initial mutual extension probability between the two matching target vascular regions;

[0022] In the matching target vascular region set, initial mutual extension probabilities of each pair of matching target vascular regions corresponding to all adjacent layers are counted, and the average initial mutual extension probability is used as the mutual extension probability of the matching target vascular region set.

[0023] Furthermore, the method for obtaining the change trend characteristics includes:

[0024] Obtaining the average respiratory influence degree of each set of matching target vascular regions;

[0025] Arranging the matching target vascular region sets in each combination in descending order based on the mutual extension probability, and the average respiratory influence degrees of the matching target vascular region sets constitute a respiratory influence degree sequence;

[0026] In the respiratory impact degree sequence, the element difference of the latter element minus the former element is obtained; the accumulated value of the element difference is used as the numerator, and the accumulated value of the absolute value of the element difference is used as the denominator to obtain the change trend characteristics of each combination method.

[0027] Furthermore, the screening method for the optimal combination includes:

[0028] Based on the change trend characteristics, the number of layers contained in the matching target vascular area set, the degree of fusiform characteristics of the target vascular area corresponding to the matching target vascular area set, and the number of elements in the combination method, the venous branching possibility of each combination method is obtained, and the combination method with the greatest venous branching possibility is selected as the optimal combination method.

[0029] Furthermore, the method for obtaining the possibility of venous branching includes:

[0030] For each combination, the maximum spindle feature degree of the target vascular area in the matching target vascular area set is obtained; the product of the maximum spindle feature degree and the number of layers is used as the venous branching feature of the matching target vascular area set; the accumulated value of the venous branching feature in the combination is multiplied by the change trend feature, and the number of matching target vascular area sets in the combination is subtracted from the product to obtain the venous branching possibility.

[0031] Furthermore, the method for obtaining the blood vessel region includes:

[0032] Edge detection is performed on each frame of CT image, and the closed area formed by the edge is regarded as the blood vessel area.

[0033] The present invention has the following beneficial effects:

[0034] The present invention analyzes the CT images at each level of the liver, taking into account that under normal circumstances, the patient's liver will produce respiratory-like state characteristics due to blood flow and physiological activities. The portal vein has weaker respiratory characteristics than the hepatic vein, and as the level changes, the respiratory influence of the portal vein will change with the position of the region. Therefore, the present invention uses the shape change and area change of the vascular region to quantify the respiratory influence. By performing matching analysis on the target vascular regions between adjacent levels, the changing trend characteristics of the respiratory influence degree are determined. The present invention uses matching and then combining to traverse all target vascular regions. Based on the changing trend characteristics, it can determine the optimal combination method to represent the portal vein, and then screen out the vascular region represented by the portal vein. Through three-dimensional reconstruction, it can achieve effective assistance in outlining the adjacent regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is a flow chart of an auxiliary method for delineating adjacent areas before liver segment resection provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0037] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a method for assisting in delineating adjacent areas prior to liver segmentectomy. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0038] Unless defined otherwise, 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 belongs.

[0039] The specific scheme of the auxiliary method for delineating adjacent areas before liver segment resection provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0040] See also Figure 1 , which shows a flow chart of an auxiliary method for delineating adjacent areas before liver segment resection provided by one embodiment of the present invention, the method comprising:

[0041] Step S1: obtaining a continuous multi-frame CT image of the patient's liver at each layer under the CT layer scan, wherein the vascular region in the CT image of the first frame is used as the target vascular region.

[0042] The embodiment of the present invention is aimed at the three-dimensional reconstruction scene of the liver portal vein after the CT layer scan has been completed. The state characteristics of each position of the liver can be identified through the CT layer scan of multiple layers. Conventional layer scans will have information on four layers. The embodiment of the present invention collects multiple frames of continuous CT images for each layer, and then transfers the images to a computer for image analysis and identification. The grayscale value of the CT image is the HU value. Different substances such as air, fat, water, soft tissue, and blood have different HU value ranges. Therefore, the vascular area can be effectively determined in the CT image for analysis of the vascular type.

[0043] It's important to note that the intrahepatic ducts can be divided into two parts: the Glisson system and the hepatic venous system. The branches and distribution of these two systems divide the liver into eight functionally independent segments, each with its own vascular inflow, outflow, and bile drainage. At the center of each segment are branches of the portal vein, hepatic artery, and bile duct. Each segment is surrounded by blood vessels draining through the hepatic vein. This means that a single CT image contains information about multiple liver segments.

[0044] Within each liver segment, the portal vein branches into the core region of each segment, forming the main blood supply for that segment. Its branching pattern is relatively constant, with little variation. For example, the right branch primarily supplies the right lobe, while the left branch covers the left lateral and left medial lobes, as well as the papillary processes. This central distribution makes the portal vein a "core landmark" for segmental demarcation, and its branching pathways form potential longitudinal fissures between segments in three-dimensional space. However, portal vein branches may interdigitate with adjacent segments, resulting in irregular demarcations. Hepatic veins (such as the right, middle, and left hepatic veins) primarily course within the longitudinal fissures between segments, draining blood from multiple adjacent segments. For example, the right hepatic vein separates the right anterior and right posterior lobes, and the middle hepatic vein divides the left and right hemilivers. Due to the high variability of the hepatic veins (e.g., dual right hepatic veins occur in 22-36% of cases), their reliability as demarcation landmarks is relatively low. Furthermore, the acute angles formed by the hepatic veins with the inferior vena cava further influence the location of the demarcation in three-dimensional space. Therefore, it is necessary to distinguish the characteristic manifestations of portal vein branches and hepatic veins in the initially divided liver segments, so as to construct a clearer three-dimensional image of the liver segment vessels.

[0045] Regarding blood vessels, contrast agents and other techniques can be used during CT image capture to create a clear brightness difference between the vascular region and the liver parenchyma region. Therefore, all vascular regions can be directly identified within each CT image. It should be noted that, because each slice in this embodiment of the present invention includes multiple consecutive CT image frames, which are used to subsequently analyze the degree of respiratory impact on a specific vascular region, the first CT image frame is selected as the baseline image, and the vascular regions within it are designated as target vascular regions. Each target vascular region is then analyzed in the subsequent steps.

[0046] Preferably, in one embodiment of the present invention, since vascular contrast agents are used to enhance vascular characteristics, edge detection can be performed directly on each frame of CT image, and the closed area formed by the edge can be used as the vascular area.

[0047] Step S2: Obtain the degree of spindle-shaped features of each vascular region; at each level, match the vascular regions in adjacent frames of CT images in sequence, and obtain the degree of respiratory influence of the target vascular region based on the area change and spindle-shaped feature change between the matching regions.

[0048] The hepatic veins and portal veins in different liver segments deform to varying degrees due to respiratory motion and intraoperative traction. Therefore, in multiple consecutive CT images within a single slice, the morphology of the vascular region changes with respiratory motion. The hepatic vein, typically located in the adjacent region of the liver segment, experiences greater deformation due to respiratory effects. The portal vein, on the other hand, experiences less deformation due to respiratory effects. Therefore, the type characteristics of a vascular region can be characterized by analyzing the degree of respiratory effects. Furthermore, considering that analyzing vascular region morphology involves changes in both area and shape, and that vascular regions exhibit distinct fusiform features in CT images, embodiments of the present invention first quantify the fusiform features of each vascular region and then analyze the changes in the degree of fusiform features and area between multiple consecutive CT images within a slice. This is because the CT image is matched sequentially, starting with the first frame, and then the adjacent frames are matched. The degree of respiratory effects on each target vascular region can be determined based on the changing characteristics of the matched regions.

[0049] Preferably, in embodiments of the present invention, considering that vascular regions have relatively distinct grayscale values due to the influence of contrast agents, and that vascular regions are closed regions at the edges of CT images, non-vascular regions may be mistakenly identified as vascular regions, the average grayscale value of the vascular region can be incorporated into the analysis of the degree of fusiform features. The minimum bounding rectangle of the vascular region is obtained, and the aspect ratio of the minimum bounding rectangle is used as the initial degree of fusiform features. The product of the average grayscale value of the vascular region and the initial degree of fusiform features is used as the degree of fusiform features. Specifically, a greater degree of fusiform features indicates that the vascular region more closely conforms to the fusiform feature, and a greater mean grayscale value indicates a greater likelihood of the region being a vein.

[0050] In one embodiment of the present invention, because the spindle-shaped feature degree encompasses both the shape and grayscale value of a region, vascular regions can be screened based on the spindle-shaped feature degree, eliminating misidentified non-vascular regions. After obtaining the screened vascular regions, analysis of the respiratory impact can be continued. The specific screening process can be configured using methods such as threshold screening, which can be freely selected by those skilled in the art and is not limited here.

[0051] Preferably, in one embodiment of the present invention, matching the blood vessel regions in adjacent CT image frames sequentially includes:

[0052] For each pair of adjacent frames at each slice, the degree of overlap between the matched vascular region in the previous CT image and the vascular region in the next CT image is obtained, and the vascular region with the highest degree of overlap is selected as the new matched vascular region; the matched vascular region in the first CT image is any target vascular region. For example, starting from a pair of adjacent frames consisting of the first and second frames, a target vascular region in the first frame is selected as the matched vascular region, and the matched vascular region of this matched vascular region in the second frame is determined. Further analysis is performed on adjacent frames consisting of the second and third frames to determine the matched vascular region of each matched vascular region in the second frame in the third frame. This process then determines all matched vascular regions and the matching relationships between these regions.

[0053] Furthermore, after determining the matching relationships between vascular regions in adjacent CT image frames, the matching relationships corresponding to a target vascular region can be sequentially formed into a set of regions. Each set of regions corresponds to a target vascular region. Within this set of regions, for each set of adjacent frames, the difference in average spindle-shaped features and average area between the matched vascular regions in the previous frame and the matched vascular regions in the next frame are obtained to determine the morphological distortion degree. The morphological distortion degrees of all sets of adjacent frames are further statistically analyzed, and the degree of respiratory influence is determined based on the average morphological distortion degree and the morphological distortion degree variance. Specifically, the greater the average morphological distortion degree and the greater the morphological distortion degree variance, the stronger the respiratory deformation characteristics of the target vascular region at that level. The greater the respiratory influence, the more likely the target vascular region is to be a hepatic vein region and the less likely it is to be a portal vein region.

[0054] In the embodiment of the present invention, since both the average morphological distortion degree and the morphological distortion degree variance are positively correlated with the respiratory influence degree, the product of the two can be directly used as the respiratory influence degree.

[0055] Furthermore, one embodiment of the present invention quantifies the degree of morphological distortion. For each set of adjacent frames, a first ratio of the average degree of spindle-shaped features and a second ratio of the average areas between the matched vascular regions of the previous frame and the matched vascular regions of the next frame are obtained. A first product of the first and second ratios is then obtained. The closer the first product is to 1, the closer both the first and second ratios are to 1, and the less distortion there is in the average degree of spindle-shaped features and average areas between the two frames. Therefore, the difference between the first product and the positive integer 1 is normalized. The greater the difference, the greater the distortion between the two frames. Thus, the degree of morphological distortion can be obtained. It should be noted that the normalization in the embodiments of the present invention can employ a linear normalization method. Those skilled in the art can select an appropriate normalization method for processing, and this will not be described in detail here.

[0056] After the processing in step S2 , each target blood vessel region in each layer has a corresponding respiratory influence degree.

[0057] Step S3: Match the target vascular regions with the same liver segment between adjacent layers. The target vascular regions with matching relationships constitute a set of matching target vascular regions. In the set of matching target vascular regions, the mutual extension probability of the set of matching target vascular regions is obtained based on the distance and position distribution between the regions.

[0058] The present invention aims to analyze the distribution of portal veins in the liver at different levels, and then obtain accurate and effective portal vein three-dimensional modeling results to assist in the delineation of adjacent areas. Therefore, it is also necessary to perform correlation analysis on the target vascular regions between adjacent levels. Therefore, step S3 further matches the target vascular regions of the same liver segment between adjacent levels, and the target vascular regions with matching relationships form a matching target vascular region set. That is, a matching target vascular region set is a set of target vascular regions at multiple levels. These target vascular regions can be regarded as local areas of a certain venous vessel. Therefore, ideally, a matching target vascular region set can correspond to a certain venous vessel. However, the target vascular region that is successfully matched is not necessarily the real same type of venous vessel region. It is necessary to further analyze its distance and position distribution and quantify the mutual extension probability in the matching target vascular region set. That is, the greater the mutual extension probability, the more likely the matching target vascular region set is to be the same type of venous vessel region.

[0059] Preferably, in an embodiment of the present invention, similar to matching adjacent CT image frames in one layer, matching target vascular regions of the same liver segment between two layers includes:

[0060] For any two adjacent layers, the first frame CT images between the two layers are matched, and a group of target vascular regions with the largest degree of overlap are selected as matching target vascular region pairs with a matching relationship; all layers are traversed, and the target vascular regions contained in the matching target vascular region pairs with a matching relationship are combined into a matching target vascular region set.

[0061] Preferably, in an embodiment of the present invention, considering that the same venous vessel regions should represent similar regional trends between different liver slices, and the matching target vessel regions between adjacent slices should have similar postures and similar distances, the method for obtaining the mutual extension probability includes:

[0062] In the set of matched target vascular regions, for two matching target vascular regions corresponding to adjacent slices, the minimum bounding rectangles of the two matching target vascular regions are obtained. The sides of the two minimum bounding rectangles are extended, and the minimum angle formed by the extended lines is used as the posture difference angle. The larger the posture difference angle, the greater the difference in posture between the two regions.

[0063] The distance between the two matching target vascular regions is obtained. Therefore, the greater the distance and the larger the posture difference angle, the less likely the two regions are the result of extensions of the same vein region between different layers. Therefore, the Euclidean norm is calculated based on the distance and the posture difference angle, and the Euclidean norm is negatively correlated to obtain the initial mutual extension probability between the two matching target vascular regions. It should be noted that in this embodiment of the present invention, the negative correlation mapping result can be expressed in inverse form. Those skilled in the art can also use other basic mathematical methods to achieve this, which will not be elaborated here.

[0064] Because multiple adjacent slices are analyzed within a set of matching target vascular regions, the initial mutual extension probability of each pair of matching target vascular regions corresponding to all adjacent slices is calculated. The average initial mutual extension probability is then used as the mutual extension probability for the set of matching target vascular regions. In other words, the greater the mutual extension probability, the more likely the set of matching target vascular regions is an extension of the same vein region.

[0065] Step S4: For each liver segment, all matching target vascular region sets contained therein are traversed and combined, and arranged according to the mutual extension probability to obtain multiple combinations; the changing trend characteristics of the respiratory influence degree of the matching target vascular region set in each combination are obtained; and the optimal combination is screened out based on the changing trend characteristics, the number of layers contained in the matching target vascular region set, the degree of the fusiform characteristics of the target vascular region corresponding to the matching target vascular region set, and the number of elements in the combination.

[0066] In step S3, matching target vascular region sets representing information at different levels are obtained. Different matching target vascular region sets can be regarded as venous regions at different locations within a liver segment. In order to obtain a complete portal vein region in a liver segment, it is necessary to determine which venous regions represented by matching vascular region sets belong to the portal vein region. Therefore, the embodiment of the present invention adopts a traversal analysis method. For each liver segment, all matching target vascular region sets contained therein are traversed and combined. For example, if four matching target vascular region sets A, B, C, and D are included, nine combinations of AB, AC, AD, BC, BD, CD, ABC, ACD, and BCD will be formed.

[0067] In CT images of multiple different layers, the diameter of portal vein branches gradually decreases as they penetrate deeper into the liver segment, and the degree to which they are affected by respiration gradually increases. The hepatic vein, on the other hand, typically has a relatively uniform diameter, so its degree of respiration is generally greater, and the values within different layers are approximately consistent. This means that the degree to which portal vein branches are affected by respiration changes as the layers extend. The mutual extension probability of the matching target vascular regions represents the degree of extension of the regions within the set. A greater degree of extension indicates that the final degree of respiratory influence of the matching target vascular regions will be amplified overall, while a lower degree of extension indicates that the degree of respiratory influence remains relatively small. In other words, for matching target vascular regions, their overall degree of respiratory influence is correlated with the mutual extension probability. Therefore, embodiments of the present invention rank the matching target vascular regions in a combination according to the mutual extension probability, and then determine the changing trend characteristics of the degree of respiratory influence. The greater the correlation between the changing trend characteristics and the arrangement characteristics of the mutual extension probability, the more consistent the combination is with portal vein characteristics.

[0068] Furthermore, the portal vein region extends across multiple liver layers and exhibits a pronounced fusiform shape. Because portal vein branches are located near the end of a liver segment, the number of portal vein branches present in that segment should be relatively small, and therefore the number of elements in the combination scheme should also be relatively small. Therefore, the optimal combination scheme can be selected based on the variation trend characteristics, the number of layers included in the matching target vascular region set, the degree of fusiform shape of the target vascular region corresponding to the matching target vascular region set, and the number of elements in the combination scheme.

[0069] Preferably, in an embodiment of the present invention, the matching target vascular region sets in each combination are arranged in descending order based on the mutual extension probability. That is, the obtained change trend feature should also be a decreasing trend feature. Therefore, the method for obtaining the change trend feature includes:

[0070] Obtaining the average respiratory influence degree of each set of matching target vascular regions;

[0071] The matching target blood vessel region sets in each combination are arranged in descending order based on the mutual extension probability, and the average respiratory influence degrees of the matching target blood vessel region sets constitute a respiratory influence degree sequence.

[0072] In the respiratory impact degree sequence, the difference between the last element and the previous element is obtained. The cumulative value of these differences is used as the numerator, and the cumulative value of the absolute values of these differences is used as the denominator to obtain the change trend characteristics of each combination. The corresponding ratio of the change trend characteristics ranges from -1 to 1. The closer the ratio is to 1, the more obvious the downward trend is and the greater the possibility of portal vein branching.

[0073] It should be noted that, in other embodiments of the present invention, if the matching target vascular region sets in each combination are arranged in ascending order of mutual extension probability, the rising trend characteristics of the average respiratory influence degree should be analyzed, which will not be described in detail.

[0074] Furthermore, the screening method for the optimal combination includes:

[0075] Based on the change trend characteristics, the number of layers included in the matching target vascular region set, the degree of the fusiform characteristics of the target vascular region corresponding to the matching target vascular region set, and the number of elements in the combination method, the venous branching possibility of each combination method is obtained, and the combination method with the greatest venous branching possibility is regarded as the optimal combination method. The method for obtaining the venous branching possibility includes:

[0076] For each combination, the maximum spindle feature degree of the target blood vessel region in each set of matching target blood vessel regions is obtained.

[0077] The product of the maximum spindle-shaped feature degree and the number of layers is used as the venous branching feature for each matching target vascular region set. Specifically, a greater number of layers indicates a closer match to the portal vein's multi-layered extension of the liver segment; a greater maximum spindle-shaped feature degree indicates a closer match to the portal vein's shape characteristics.

[0078] Because a single combination has multiple matching target vascular region sets, the cumulative value of the venous branching features in the combination is multiplied by the change trend feature, and the number of matching target vascular region sets in the combination is subtracted from the product to obtain the venous branching probability. Specifically, the greater the change trend feature, the more consistent the combination is with the change characteristics of the portal vein region; the greater the cumulative value of the venous branching features, the more consistent the matching target vascular region sets in the combination are with the characteristics of portal vein branch regions; and the smaller the number of matching target vascular region sets in the combination, the more consistent it is with the distribution characteristics of the portal vein within the liver segment. Therefore, the greater the final quantitative venous branching probability, the more likely the combination is to be a portal vein branch combination.

[0079] Step S5: The target vascular region corresponding to the optimal combination is used as a portal vein branch and three-dimensionally reconstructed to assist in delineating the adjacent region.

[0080] Once the optimal combination for representing portal vein branches is determined, three-dimensional reconstruction can be performed based on the corresponding target vascular region as the portal vein branch region. Three-dimensional reconstruction can be performed using the CT machine's built-in post-processing workstation. Reconstruction methods include multiplanar reformation (MPR), volume rendering (VR), and maximum intensity projection (MIP). MPR reconstruction includes coronal, sagittal, and curved planar reformation (CPR) views, with a slice thickness of 2.0 mm, an interval of 2.0 mm, a window width of 250 Hu, and a window position of 50 Hu.

[0081] For all liver segments, reconstructed images of the corresponding portal vein branches can be obtained; based on the portal vein branch segmentation results on the above liver segments, doctors can be assisted in outlining the adjacent areas between different liver segments, thereby improving the accuracy of the liver segmentation process.

[0082] In summary, the embodiment of the present invention analyzes the CT images at each level of the liver, and uses the shape changes and area changes of the vascular region to quantify the degree of respiratory influence. By performing matching analysis on the target vascular regions between adjacent levels, the changing trend characteristics of the degree of respiratory influence are determined, and then all target vascular regions are traversed by matching and then recombining. Based on the changing trend characteristics, the optimal combination method representing the portal vein can be determined, and the vascular region represented by the portal vein can be screened out, and effective assistance in outlining adjacent regions can be achieved through three-dimensional reconstruction. The present invention can effectively assist in outlining adjacent regions by analyzing the correlation of vascular regions in CT images at different levels, screening out accurate portal vein regions, and performing three-dimensional reconstruction.

[0083] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for assisting the delineation of adjacent areas before liver segment resection, characterized in that: The method comprises: Obtaining multiple consecutive CT images of the patient's liver at each layer of the CT scan, wherein the vascular region in the first frame of the CT image is used as the target vascular region; Obtain the degree of spindle-shaped features of each vascular region; at each layer, the vascular regions in adjacent CT frames are matched sequentially, and the degree of respiratory influence of the target vascular region is obtained based on the area change and spindle-shaped feature change between the matched regions; Matching target vascular regions of the same liver segment between adjacent slices, forming a set of matching target vascular regions with matching relationships; obtaining a mutual extension probability of the set of matching target vascular regions based on the distance and position distribution between the regions; For each liver segment, all matching target vascular region sets contained therein are traversed and combined, and arranged according to mutual extension probability to obtain multiple combinations. The changing trend characteristics of the respiratory influence degree of the matching target vascular region set in each combination are obtained. The optimal combination is screened based on the changing trend characteristics, the number of layers contained in the matching target vascular region set, the degree of fusiform characteristics of the target vascular region corresponding to the matching target vascular region set, and the number of elements in the combination. The target vascular area corresponding to the optimal combination is used as the portal vein branch and three-dimensionally reconstructed to assist in delineating the adjacent area; The method for obtaining the degree of the spindle-shaped feature includes: Obtaining a minimum circumscribed rectangle of the blood vessel region, using the aspect ratio of the minimum circumscribed rectangle as an initial spindle-shaped feature degree, and multiplying the average grayscale value of the blood vessel region by the initial spindle-shaped feature degree as the spindle-shaped feature degree; Methods for obtaining mutual extension probability include: In the set of matching target vascular regions, for two matching target vascular regions corresponding to adjacent slices, obtaining the minimum circumscribed rectangles of the two matching target vascular regions, extending the sides of the two minimum circumscribed rectangles, and using the minimum angle formed by the extended lines as the posture difference angle; obtaining a distance between the two matching target vascular regions, calculating a Euclidean norm based on the distance and the posture difference angle, performing negative correlation mapping on the Euclidean norm, and obtaining an initial mutual extension probability between the two matching target vascular regions; In the matching target vascular region set, initial mutual extension probabilities of each pair of matching target vascular regions corresponding to all adjacent layers are counted, and the average initial mutual extension probability is used as the mutual extension probability of the matching target vascular region set.

2. The method for assisting in delineating adjacent areas before liver segment resection according to claim 1, characterized in that: Matching the blood vessel regions in adjacent CT images in sequence includes: For each pair of adjacent frames at each layer, the degree of overlap between the matched vascular region in the previous CT image and the vascular region in the next CT image is obtained, and the vascular region with the highest overlap is selected as the new matched vascular region; the matched vascular region in the first CT image is any target vascular region; Match target vascular regions that share the same liver segment between adjacent slices, including: For any two adjacent layers, the first frame CT images between the two layers are matched, and a group of target vascular regions with the largest degree of overlap are selected as matching target vascular region pairs with a matching relationship; all layers are traversed, and the target vascular regions contained in the matching target vascular region pairs with a matching relationship are combined into a matching target vascular region set.

3. The method for assisting in delineating adjacent areas before liver segment resection according to claim 2, characterized in that: The method for obtaining the respiratory impact degree includes: For each group of adjacent frames, the difference in the average degree of spindle-shaped features and the difference in average area between the matched vascular regions in the previous frame and the next frame are obtained to obtain the morphological distortion degree; the morphological distortion degrees of all groups of adjacent frames are counted, and the degree of respiratory influence is obtained based on the average morphological distortion degree and the morphological distortion degree variance.

4. The method for assisting in delineating adjacent areas before liver segment resection according to claim 3, characterized in that: The method for obtaining the morphological distortion degree includes: For each group of adjacent frames, a first ratio of the average degree of spindle-shaped features between the matched blood vessel region in the previous frame and the matched blood vessel region in the next frame, and a second ratio of the average areas are obtained; a first product between the first ratio and the second ratio is obtained, and the difference between the first product and the positive integer 1 is normalized to obtain the morphological distortion degree.

5. The method for assisting in delineating adjacent areas before liver segment resection according to claim 1, characterized in that: The method for obtaining the change trend feature includes: Obtaining the average respiratory influence degree of each set of matching target vascular regions; Arranging the matching target vascular region sets in each combination in descending order based on the mutual extension probability, and the average respiratory influence degrees of the matching target vascular region sets constitute a respiratory influence degree sequence; In the respiratory impact degree sequence, the element difference of the latter element minus the former element is obtained; the accumulated value of the element difference is used as the numerator, and the accumulated value of the absolute value of the element difference is used as the denominator to obtain the change trend characteristics of each combination method.

6. The method for assisting in delineating adjacent areas before liver segment resection according to claim 5, characterized in that: The screening method of the optimal combination includes: Based on the change trend characteristics, the number of layers contained in the matching target vascular area set, the degree of fusiform characteristics of the target vascular area corresponding to the matching target vascular area set, and the number of elements in the combination method, the venous branching possibility of each combination method is obtained, and the combination method with the greatest venous branching possibility is selected as the optimal combination method.

7. The method for assisting in delineating adjacent areas before liver segment resection according to claim 6, characterized in that: The method for obtaining the possibility of venous branching includes: For each combination, the maximum spindle feature degree of the target vascular area in the matching target vascular area set is obtained; the product of the maximum spindle feature degree and the number of layers is used as the venous branching feature of the matching target vascular area set; the accumulated value of the venous branching feature in the combination is multiplied by the change trend feature, and the number of matching target vascular area sets in the combination is subtracted from the product to obtain the venous branching possibility.

8. The method for assisting in delineating adjacent areas before liver segment resection according to claim 1, characterized in that: The method for obtaining the blood vessel region includes: Edge detection is performed on each frame of CT image, and the closed area formed by the edge is regarded as the blood vessel area.

Images