Method and system for corruption interference partition correction for virtual dissection gas imagery

Abstract

The application discloses a corruption interference partition correction method and system for virtual dissection gas images, and belongs to the technical field of image interference correction. The method specifically comprises the following steps: acquiring three-dimensional virtual dissection image data, registering the three-dimensional virtual dissection image data to a preset dissection standard space, constructing a candidate gas topology graph, and associating the candidate gas topology graph with a corruption correlation parameter; calculating a corresponding corruption interference potential value; according to the spatial propagation gradient of the corruption interference potential value in the candidate gas topology graph, dividing candidate gas voxels into a plurality of corruption interference partitions; respectively correcting the plurality of corruption interference partitions of the candidate gas voxels in a differentiated manner; fusing the partition correction results after the differentiated partition correction with original three-dimensional virtual dissection image data to generate virtual dissection gas images after the corruption interference correction; and the application reduces the interference of corruption-related low gray level signals and reduces the situation that diagnostic gas features are mistakenly deleted or corruption gas is mistakenly retained.

Description

Technical Field

[0001] This invention belongs to the field of image interference correction technology, specifically a method and system for zoning correction of putrefaction interference in virtual anatomical gas images. Background Technology

[0002] Virtual anatomy is a technique that uses computed tomography, magnetic resonance imaging, or multimodal 3D image reconstruction to perform non-invasive observation and analysis of the internal structure of an object under examination. In virtual anatomy imaging, gas images often have significant interpretive value; for example, they can be used to assist in identifying imaging features such as gas accumulation in cavities, gas within lumens, interstitial gas, gas associated with perforation paths, and localized abnormal cavities.

[0003] In virtual autopsy scenarios, gas images do not entirely originate from the event being analyzed. With variations in factors such as detachment intervals, ambient temperature, humidity, storage conditions, cooling records, and transport processes, the putrefaction process may generate a large number of low-density gas signals in different regions, which can then migrate and diffuse along cavities, tubes, fissures, or interlayer spaces. Existing virtual autopsy gas image processing methods often rely on fixed grayscale thresholds, manual interpretation based on experience, or simple region segmentation methods. These methods typically only identify low-grayscale gas voxels themselves, making it difficult to further distinguish between in-situ putrefaction generation, putrefaction migration and diffusion, and diagnostic gas regions that should be preserved. This leads to problems such as gas artifacts, boundary confusion, misjudgment of connectivity, and difficulty in identifying gas sources in virtual autopsy images. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method and system for zoning correction of putrefaction interference in virtual anatomical gas images. By combining three-dimensional spatial topology, putrefaction correlation parameters, gas boundary states, and propagation gradient relationships, the method performs zoning identification and differentiated correction of putrefaction interference in virtual anatomical gas images, thereby improving the pertinence of the correction process and the usability of the image results.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for zoning correction of putrefaction interference in virtual anatomical gas imaging includes: Acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology map containing candidate gas voxels, connected edges and spatial adjacency relationships based on gas grayscale threshold, voxel connectivity and gas boundary gradient. Based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency between them, and combined with the putrefaction correlation parameters, the corresponding putrefaction interference potential is calculated. According to the spatial propagation gradient of the putrefaction interference potential in the candidate gas topology map, the candidate gas voxel is divided into multiple putrefaction interference zones, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone. Differential corrections were performed on multiple corruption interference zones of the candidate gas voxels; The partition correction results after partition differentiation correction are fused with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

[0006] Specifically, the step of acquiring three-dimensional virtual anatomical image data, registering the three-dimensional virtual anatomical image data to a preset anatomical standard space, and constructing a candidate gas topology map including candidate gas voxels, connected edges, and spatial adjacency relationships includes: Acquire three-dimensional virtual anatomical image data, and perform voxel scale unification, gray level normalization and background voxel removal processing on the three-dimensional virtual anatomical image data to generate three-dimensional image data to be registered. From the three-dimensional image data to be registered, extract bony anatomical anchor points, body surface contour anchor points, and cavity contour anchor points. Determine the overall posture and shape boundary based on the bony anatomical anchor points, and map the three-dimensional image data to be registered to a preset anatomical standard space to obtain standardized three-dimensional image data. Based on the standardized three-dimensional image data, multiple anatomical regions are determined, and corresponding gas grayscale filtering ranges are configured for different anatomical regions. Voxels in each anatomical region are gated according to the gas grayscale filtering range to generate an initial low grayscale voxel set. Neighborhood grayscale detection is performed on the initial low grayscale voxels in the initial low grayscale voxel set to construct a gas boundary shell around the initial low grayscale voxels. The initial low grayscale voxels that meet the gas boundary shell existence condition and the preset voxel adjacency condition are marked as candidate gas voxels. Based on the spatial adjacency relationship between the candidate gas voxels, connected edges are established, and disconnection markers are used as blocking conditions for connected edges. The anatomical region where the candidate gas voxel is located, the gas boundary shell type, the adjacency direction, and the connection scale are written into the attribute items of the corresponding connected edge to construct a candidate gas topology graph containing candidate gas voxels, connected edges, and spatial adjacency relationships.

[0007] Specifically, based on the standardized 3D image data, multiple anatomical regions are determined, and corresponding gas grayscale filtering ranges are configured for different anatomical regions. Voxels within each anatomical region are then gated according to the gas grayscale filtering ranges to generate an initial set of low-grayscale voxels, including: According to the spatial coordinates, density levels and closed contours in the preset anatomical standard space, the standardized three-dimensional image data is divided into cavity domain, canal domain, solid domain, surface domain and fissure domain, and corresponding partition markers are written for each voxel. Within the region corresponding to each partition marker, low-density voxels, transition-density voxels, and adjacent high-density voxels are selected respectively to form a partition grayscale sample set, and a grayscale level sequence corresponding to the partition is generated based on the partition grayscale sample set. Based on the grayscale level sequence, spatial enclosure degree and neighborhood density difference of each partition, the corresponding gas grayscale screening range is configured respectively. Specifically, the first grayscale screening range is configured for the cavity domain, the second grayscale screening range is configured for the pipe domain, the third grayscale screening range is configured for the solid domain, and the fourth grayscale screening range is configured for the surface domain and the fissure domain. According to the partition label of each voxel, the corresponding gas grayscale filtering range is called, and the voxels in each partition are gated and filtered; the voxels that fall into the corresponding gas grayscale filtering range are written into the initial low grayscale voxel set, and the partition label, grayscale level label and gate source label of the voxel are written into the initial low grayscale voxel set.

[0008] Specifically, neighborhood grayscale detection is performed on the initial low-grayscale voxels in the initial low-grayscale voxel set to construct a gas boundary shell around the initial low-grayscale voxels. Initial low-grayscale voxels that satisfy the gas boundary shell existence condition and the preset voxel adjacency condition are marked as candidate gas voxels, including: Based on the partition label, gray level label, and gate source label of each initial low gray voxel in the initial low gray voxel set, a corresponding neighborhood sampling window is set, and low gray voxels, transition gray voxels, and non-low gray voxels adjacent to the initial low gray voxel are read within the neighborhood sampling window. Starting with an initial low-grayscale voxel, the grayscale change sequence of neighboring voxels is read along multiple preset spatial directions. The voxel sequence that continuously transitions from low-grayscale voxels to non-low-grayscale voxels is marked as a boundary transition zone, and the spatial direction and grayscale level change mark corresponding to the boundary transition zone are recorded. Multiple boundary transition zones surrounding the same initial low-grayscale voxel are closed and matched according to the spatial direction. When multiple boundary transition zones form an outer envelope that satisfies the preset closure condition on the outer periphery of the initial low-grayscale voxel, the outer envelope is determined as the gas boundary shell corresponding to the initial low-grayscale voxel. The preset closure condition is used to determine whether an envelope structure that can characterize the boundary of a gas region is formed around a low-grayscale voxel. For cavity-like or cluster-like low-grayscale regions, if the initial low-grayscale voxel has a continuous sequence of changes from low-grayscale voxels, transitional grayscale voxels to non-low-grayscale voxels in at least three non-coplanar spatial directions, and these boundary transition zones can form a continuous or nearly continuous outer envelope around the low-grayscale region, then it is determined that it meets the preset closure condition. For tubular or fissure-like low-grayscale regions, if they remain continuous in the main extension direction and form boundary transition zones on both sides or multiple sides of the cross section perpendicular to the main extension direction, then they are determined to meet the corresponding semi-closure or cross-sectional closure condition. Low-grayscale voxels that only extend in a single direction, lack stable transitional grayscale voxels around the periphery, or are directly connected to the background region are not considered to meet the preset closure condition.

[0009] Initial low-grayscale voxels with gas boundary shells and satisfying preset voxel adjacency conditions are marked as candidate gas voxels, and initial low-grayscale voxels that do not form gas boundary shells or are only adjacent through unidirectional narrow-diameter low-grayscale channels are written into the candidate exclusion set.

[0010] Specifically, based on each candidate gas voxel in the candidate gas topology map and their spatial adjacency relationships, and combined with the putrefaction correlation parameters, the corresponding putrefaction interference potential is calculated. According to the spatial propagation gradient of the putrefaction interference potential in the candidate gas topology map, the candidate gas voxels are divided into multiple putrefaction interference partitions, including: Obtain putrefaction correlation parameters corresponding to the three-dimensional virtual anatomical image data, map the putrefaction correlation parameters to a preset anatomical standard space, and write corresponding external putrefaction markers for candidate gas voxels in the candidate gas topology map. The putrefaction correlation parameters include the in vitro interval, storage temperature, storage humidity, sealing status, cooling record, and pre-collection transport status. Based on each candidate gas voxel in the candidate gas topology map, and in conjunction with the external putrefaction marker, a local putrefaction attribute label for the candidate gas voxel is generated. According to the local corrosion attribute label, each candidate gas voxel is assigned a corrosion interference potential value. Among them, candidate gas voxels located in closed cavity domains, fissure domains, or surface domains and distributed in a cluster are assigned a first type of corrosion interference potential value. Candidate gas voxels that extend in a chain along the pipe domain or interlayer gap are assigned a second type of corrosion interference potential value. Candidate gas voxels adjacent to sudden fracture boundaries, penetration paths, or local abnormal cavities are assigned a third type of corrosion interference potential value. Using the connected edges in the candidate gas topology graph as the propagation path, the propagation order of the corruption interference potential in the candidate gas topology graph is determined based on the corruption interference potential, adjacency direction and connectivity scale between adjacent candidate gas voxels, and the corresponding spatial propagation gradient is formed. Candidate gas voxels are partitioned and labeled according to the spatial propagation gradient, including in-situ putrefaction generation region, putrefaction migration and diffusion region, and diagnostic gas retention region. Candidate gas voxels located at the cluster center and whose putrefaction interference potential values ​​are transmitted in the same way between adjacent nodes are labeled as in-situ putrefaction generation region. Candidate gas voxels extending outward from the cluster center along the spatial propagation gradient are labeled as putrefaction migration and diffusion region. Candidate gas voxels that are disconnected from the spatial propagation gradient and have a third type of putrefaction interference potential value are labeled as diagnostic gas retention region.

[0011] Specifically, the process of using connected edges in the candidate gas topology graph as propagation paths, determining the propagation order of the corruption interference potential in the candidate gas topology graph based on the corruption interference potential, adjacency direction, and connectivity scale between adjacent candidate gas voxels, and forming the corresponding spatial propagation gradient includes: Read the corruption interference potential, adjacency direction and connectivity scale of the candidate gas voxels at both ends of each connected edge in the candidate gas topology graph, and mark the connected edges as propagable edges or blocking edges according to the continuity of the gas boundary shell and the cross-region situation. In the candidate gas topology graph, candidate gas voxels whose corruption interference potential values ​​are clustered in the same type and whose number of connected edges meets the preset clustering conditions are determined as the propagation starting points. When there are multiple propagation starting points, the order of propagation starting points is determined according to the partition, connectivity scale and gas boundary shell state. Taking the propagation starting point as the initial node, read the adjacent candidate gas voxels along the propagable edge, and determine the propagation order of the corruption interference potential value based on the continuity relationship of the corruption interference potential value, the consistency of the adjacent direction, and the connection relationship of the connectivity scale between the adjacent candidate gas voxels. According to the propagation order, candidate gas voxels are divided into origin layer, transition layer, extension layer and termination layer, and candidate gas voxels isolated by blocking edges are divided into independent levels. The origin layer, transition layer, extension layer, termination layer and independent levels form a corresponding spatial propagation gradient.

[0012] Specifically, the step of partitioning and labeling candidate gas voxels based on spatial propagation gradients includes: Based on the spatial propagation gradient and the corruption interference potential of candidate gas voxels, the connected edges and the gas boundary shell type, a voxel determination sequence is generated. In the voxel determination sequence, candidate gas voxels located in the origin layer and continuously adjacent to voxels of the same type of putrefaction interference potential are marked as in-situ putrefaction generation regions. For candidate gas voxels that are clustered in a ring, cluster or island shape around the same propagation starting point, they are incorporated into the same in-situ putrefaction generation region. Along the spatial propagation gradient from the origin layer to the transition layer, extension layer and termination layer, candidate gas voxels connected to the in-situ putrefaction generation zone through a propagable edge are read sequentially. Candidate gas voxels whose putrefaction interference potential has a continuous relationship and whose adjacent direction is consistent with the propagation direction are marked as putrefaction migration and diffusion zones. Candidate gas voxels located at independent levels or isolated by blocked edges are read, and it is determined whether they form a propagation disconnect relationship with the in-situ putrefaction generation zone and putrefaction migration and diffusion zone. Candidate gas voxels that form a propagation disconnect relationship and have an isolated gas state type are marked as diagnostic gas retention zones.

[0013] Specifically, the differential correction of multiple corruption interference regions of the candidate gas voxels includes: Read the putrefaction interference partition marker, putrefaction interference potential and spatial propagation gradient level of the candidate gas voxel, and match the candidate gas voxel to the in-situ replacement strategy, migration backoff strategy or retention lock strategy according to the putrefaction interference partition marker. For candidate gas voxels marked as in-situ putrefaction generation zones, the gray levels of neighboring non-gas voxels are read, and a virtual filling value is generated by combining the reference gray level sequence of the partition to which the candidate gas voxel belongs. The virtual filling value is then written to the corresponding candidate gas voxel position. For candidate gas voxels marked as putrefaction migration and diffusion zones, traversal is performed in reverse order of spatial propagation gradient, and the gas signals in the putrefaction migration and diffusion zones are segmented and backtracked based on the adjacency direction and connectivity scale between adjacent candidate gas voxels. For candidate gas voxels marked as diagnostic gas retention areas, their independent levels and blocking edge markers are read; when there is a blocking edge between the candidate gas voxel and the in-situ putrefaction generation area or putrefaction migration and diffusion area, a retention lock marker is set for the candidate gas voxel. At the boundary of different corruption interference zones, the correction assignment of boundary voxels is determined based on the zone marker, boundary shell type, and spatial propagation gradient level. Candidate gas voxels that have undergone in-situ replacement, migration back and retention locking are then merged into the zone correction results.

[0014] Specifically, the process of fusing the partition correction results (after partition differentiation correction) with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction includes: Read the candidate gas voxels in the partition correction results after in-situ replacement, migration rollback and retention locking, and according to the position of the candidate gas voxels in the preset anatomical standard space, map them to the original voxel positions in the original three-dimensional virtual anatomical image data to generate a set of voxels to be fused and corrected. Based on the original three-dimensional virtual anatomical image data, the original voxel grayscale that was not written into the set of voxels to be fused and corrected is retained. For the original voxels that were written into the set of voxels to be fused and corrected, the in-situ replacement value, migration rollback value or retention lock value is called according to their correction source mark to form layered write-back image data. Read the boundary position between the corrected voxels and the uncorrected voxels in the layered write-back image data, and determine the gray-level connection order of the boundary voxels according to the partition marker, boundary shell type and spatial propagation gradient level corresponding to the boundary voxels. Then, continuously reconstruct the boundary position according to the gray-level connection order to generate gas-corrected fused image data. The gas-corrected fused image data is remapped to the original image coordinate space, and the partition markers, correction source markers, and retention lock markers of each candidate gas voxel are written simultaneously to generate a virtual anatomical gas image after putrefaction interference correction.

[0015] A system for zoning correction of putrefaction interference in virtual anatomical gas images includes: a topology map construction module, a zoning module, a correction module, and a fusion module; The topology graph construction module is used to acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology graph containing candidate gas voxels, connected edges, and spatial adjacency relationships. The partitioning module is used to calculate the corresponding putrefaction interference potential value based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency relationship between them, and in combination with putrefaction correlation parameters. Based on the spatial propagation gradient of the putrefaction interference potential value in the candidate gas topology map, the candidate gas voxel is divided into multiple putrefaction interference partitions, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone. The correction module is used to perform differential correction on multiple corruption interference zones of the candidate gas voxels respectively. The fusion module is used to fuse the partition correction results after partition differentiation correction with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes a method and system for zoning correction of putrefaction interference in virtual anatomical gas images. It jointly processes gas grayscale features, voxel adjacency relationships, boundary shell states, putrefaction-related parameters, and spatial propagation gradients. First, it calculates the putrefaction interference potential value for candidate gas voxels, then divides them into in-situ putrefaction generation zones, putrefaction migration and diffusion zones, and diagnostic gas retention zones. In-situ replacement, migration rollback, and retention locking corrections are then performed for each zone. This avoids the confusion of gas sources caused by relying solely on fixed grayscale thresholds or simple connected region identification, reduces the interference of putrefaction-related low-grayscale signals on virtual anatomical gas images, and minimizes the possibility of erroneous deletion of diagnostic gas features or erroneous retention of putrefaction gases. Attached Figure Description

[0017] Figure 1 Flowchart of the putrefaction interference zoning correction method for virtual anatomical gas images provided by the present invention; Figure 2 This is a schematic diagram of the candidate gas topology provided by the present invention; Figure 3 A schematic diagram of the corruption interference partition marking provided by the present invention; Figure 4 This is a diagram illustrating the system architecture for zoning correction of putrefaction interference in virtual anatomical gas images provided by the present invention. Detailed Implementation

[0018] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0021] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0022] Example 1: Please see Figures 1-3 The present invention provides an embodiment of a method for zoning correction of putrefaction interference in virtual anatomical gas images, comprising the following specific steps: Step S1: Acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology map containing candidate gas voxels, connected edges and spatial adjacency relationships based on gas grayscale threshold, voxel connectivity and gas boundary gradient.

[0023] like Figure 1 As shown, the specific steps of step S1 are as follows: Step S101: Acquire three-dimensional virtual anatomical image data, perform voxel scale unification, gray level normalization and background voxel removal processing on the three-dimensional virtual anatomical image data, and generate three-dimensional image data to be registered.

[0024] In this embodiment, the three-dimensional virtual anatomical image data can be tomographic sequence data obtained by computed tomography or three-dimensional volume data after format conversion. In specific implementation, the slice thickness, interslice spacing, pixel spacing, scan matrix, gray level depth, and scan direction information in the image data are first read, and the actual size of the original voxels in the three spatial directions is determined according to the slice thickness, interslice spacing, and pixel spacing. When the scale of the original voxels is inconsistent in different directions, a three-dimensional resampling method is used to convert them into a uniform voxel scale, for example, converting anisotropic voxels into isotropic voxels. During the resampling process, linear interpolation or cubic interpolation can be used for continuous grayscale images, and nearest neighbor interpolation can be used for marked regions or discrete mask data. After this processing, the original tomographic image is organized into three-dimensional image data with a uniform voxel scale.

[0025] Furthermore, the 3D image data after voxel-scale unification is subjected to grayscale level normalization. Specifically, when the 3D image data comes from computed tomography, the original pixel values ​​are converted into corresponding physical grayscale values ​​based on the rescaling slope and rescaling intercept in the image file. When different scanning devices or different scanning protocols cause grayscale range shifts, air, soft tissue sample areas, high-density bone areas, or background areas outside the scanning bed are selected as grayscale references to map the grayscale levels of the entire 3D image to a preset grayscale range (0-255). For images with significant local noise, median filtering, nonlocal mean filtering, or anisotropic diffusion filtering are used to suppress noise before and after normalization, but the main grayscale differences between low-grayscale gas areas and adjacent tissue grayscale areas are not changed.

[0026] Furthermore, background voxel removal is performed on the grayscale-normalized 3D image data. First, the background region outside the scanning field of view is separated from the 3D image data. Then, based on the 3D connected component analysis, voxels such as the scanning bed, air background, edge noise blocks, and isolated artifacts that are not connected to the main region are removed. For supports or fixed materials that are adjacent to the main region but do not belong to the object to be analyzed, they are excluded based on their location, connectivity range, and grayscale distribution. After background removal, effective voxels within the spatial range of the object to be analyzed are retained. The 3D coordinates, grayscale values, voxel scales, and original image index relationships of the retained voxels are recorded, thereby generating the 3D image data to be registered.

[0027] Step S102: Extract bony anatomical anchor points, body surface contour anchor points, and cavity contour anchor points from the three-dimensional image data to be registered. Determine the overall posture and shape boundary based on the bony anatomical anchor points, and map the three-dimensional image data to be registered to a preset anatomical standard space to obtain standardized three-dimensional image data.

[0028] In this embodiment, when extracting bony anatomical anchor points from the three-dimensional image data to be registered, threshold segmentation is performed on the three-dimensional image data to be registered based on the high-density grayscale range to obtain a set of bony high-density voxels. The high-density grayscale range is obtained by finding the high grayscale peak corresponding to the bone in the image grayscale histogram and setting the grayscale interval near the peak as the high-density grayscale range. Then, three-dimensional connected component analysis, hole filling, and small artifact removal are performed on the set of bony high-density voxels to obtain candidate regions of bony structures. Specifically, anchor points with stable spatial positions are extracted from the candidate regions of bony structures, such as the centerline point of the axial bony region, the endpoints of the bilaterally symmetrical high-density structures, the axis points of the continuous long strip high-density structures, and the geometric center points of the joint-like adjacent regions. For cases of local defects or abnormal scanning posture, principal component analysis is used to determine the long axis direction, short axis direction, and lateral distribution direction of the candidate regions of bony structures, and the long axis direction is used as the main reference direction for overall posture correction. The bony anatomical anchor points obtained in this way include spatial coordinates, the label of the bony region to which they belong, and directional attributes.

[0029] Furthermore, the body surface contour anchor points are obtained by extracting the outer contour of the effective voxel region in the 3D image data to be registered. In the 3D image data after background voxel removal, the outer boundaries of the effective voxels are searched along multiple projection directions, and holes, gaps, and discrete noise points in the outer contour are removed by combining 3D morphological closing operations to form a continuous outer shape envelope. Then, the front and back boundary points, left and right boundary points, top and bottom boundary points, and contour points with large curvature changes are selected from this outer shape envelope as body surface contour anchor points. The cavity contour anchor points are obtained based on the extraction of low grayscale or medium-low grayscale connected regions. Candidate regions located inside the outer shape envelope and distributed in a tubular, cavity, or continuous void shape are first screened out, and then cavity contour anchor points are extracted based on their center line, cross-sectional contour, and end position. The body surface contour anchor points are mainly used to limit the outer shape boundary and overall scale, while the cavity contour anchor points are mainly used to constrain the internal spatial orientation and local registration deviation.

[0030] Furthermore, based on bony anatomical anchor points, body surface contour anchor points, and cavity contour anchor points, the three-dimensional image data to be registered is mapped to a preset anatomical standard space. Rigid registration transformation parameters are calculated based on the bony anatomical anchor points to ensure that the overall posture of the three-dimensional image data to be registered is consistent with the standard posture in the preset anatomical standard space. Subsequently, scale and shape boundary correction parameters are calculated based on the body surface contour anchor points to ensure that the shape envelope falls within the corresponding range of the preset anatomical standard space. Finally, local non-rigid corrections are performed based on the cavity contour anchor points to match the internal cavity orientation with the cavity reference area in the preset anatomical standard space. After completing the above mapping, the three-dimensional image data to be registered is resampled, and the index correspondence between each voxel in the original image coordinate space and the preset anatomical standard space is retained to obtain standardized three-dimensional image data. The preset anatomical standard space is a pre-established standard three-dimensional human anatomy template. Standard three-dimensional anatomical images are selected, voxel sizes and human postures are unified, and the positions of bones, body surface, and major cavities are marked. The template is then saved as a standard template, specifying unified coordinate directions, voxel sizes, human postures, and the positions of major anatomical structures.

[0031] Step S103: Based on the standardized three-dimensional image data, determine multiple anatomical partitions, configure corresponding gas grayscale filtering ranges for different anatomical partitions, and perform gating filtering on voxels in each anatomical partition according to the gas grayscale filtering range to generate an initial low grayscale voxel set.

[0032] In this embodiment, when partitioning based on standardized 3D image data, the image voxels are first spatially classified using the coordinate range, outer envelope, internal low-density continuous region, and high-density support region in a preset anatomical standard space. Specifically, regions within a closed space with continuous outer boundaries and low-density distribution are classified as cavity regions; regions extending along a fixed direction with relatively continuous cross-sections and a slender, connected shape are classified as tube regions; regions with relatively continuous grayscale distribution, a low proportion of low-density voxels, and blocky spatial occupancy are classified as solid regions; regions near the outer envelope surface and with layered distribution characteristics are classified as surface regions; and regions located between adjacent tissue layers with a narrow slit-like or irregular interlayer shape are classified as fissure regions. After completing the above partitioning, a corresponding partition marker is written for each voxel in the standardized 3D image data; for voxels located at the boundary between two partitions, a unique partition marker is determined based on their distance from the center region of the adjacent partition, grayscale continuity, and local connectivity direction, thereby obtaining partitioned image data with spatial attribution information.

[0033] Furthermore, a partitioned grayscale sample set is constructed within the region corresponding to each partition marker. Low-density voxels, transition-density voxels, and adjacent high-density voxels are selected as samples within each partition. Low-density voxels are obtained through initial screening in low grayscale ranges, transition-density voxels are extracted from a certain neighborhood around the low-density voxels, and adjacent high-density voxels are extracted from regions adjacent to the low-density voxels but with significantly increased grayscale. For cavity and tube domains, samples are preferentially selected from continuous low-density regions and their boundaries. For solid domains, samples are preferentially selected from locally stable grayscale regions, locally low-grayscale spots, and their surrounding regions. For surface and fissure domains, samples are preferentially selected from locations close to the outer envelope or interlayer gaps. The above sample selection is achieved using existing grayscale histogram statistics, local threshold segmentation, connected component screening, and neighborhood sampling techniques to form a partitioned grayscale sample set that can reflect the differences in grayscale levels within each partition.

[0034] Furthermore, a grayscale level sequence for each partition is generated based on the partitioned grayscale sample set, and the gas grayscale filtering range is configured accordingly. Specifically, low-density voxels, transitional-density voxels, and adjacent high-density voxels within each partition are sorted from low to high grayscale to form the grayscale level sequence for that partition. At the same time, the spatial closure degree, boundary continuity, grayscale differences in neighboring areas, and local connectivity of the partition are read as auxiliary criteria for configuring the gas grayscale filtering range. For cavities, a first grayscale screening range is configured to cover the low-density gas manifestation within the cavity, as the internal low-density region may have a large continuous space. For tubes, a second grayscale screening range is configured based on the tube orientation and cross-sectional continuity to distinguish between true low-grayscale channels and adjacent pseudo-low-grayscale regions within the tube. For the solid domain, a third grayscale screening range is configured to focus more on identifying local low-grayscale anomalous voxels. For the surface and fracture domains, a fourth grayscale screening range is configured, and the screening targets are limited based on the characteristics of the peripheral boundary and interlayer gaps. Each grayscale screening range does not use a globally uniform threshold, but is determined separately according to the grayscale level sequence and spatial attributes within different partitions. Specifically, for each anatomical partition, a partition grayscale range is first established based on the grayscale distribution of low-density voxels, transitional-density voxels, and adjacent high-density voxels within that partition. A histogram is generated to identify low-density gray-scale peaks, transitional gray-scale intervals, and non-gas-organic gray-scale intervals. The interval containing the low-density gray-scale peak is then used as the lower limit reference for candidate gas gray-scale values, and the gray-scale boundary between the low-density gray-scale peak and the transitional gray-scale interval is used as the upper limit reference for candidate gas gray-scale values. For cavities, since their internal gas regions typically have large continuous spaces, the first gray-scale screening range allows coverage of a wider low-density gray-scale interval. For tubular regions, the second gray-scale screening range, in addition to meeting the low-density gray-scale condition, also requires that the corresponding voxels have a continuous distribution trend along the tubular direction. For solid regions, the third gray-scale screening range is relatively narrowed, and noise-type low-gray-scale points are excluded based on local neighborhood density differences. For surface and fracture regions, the fourth gray-scale screening range is determined by combining the outer envelope distance, interlayer gap direction, and boundary continuity.

[0035] Finally, according to the partition markers already written to the voxels, the corresponding gas grayscale filtering ranges are called zone by zone to perform gated filtering on the standardized 3D image data. First, the partition marker of the target voxel is read, and then the first, second, third, or fourth grayscale filtering range corresponding to that partition is called to determine whether the voxel's grayscale falls into the corresponding range. If it falls into the corresponding range, the voxel is written into the initial low grayscale voxel set. If it does not fall into the corresponding range, it remains a non-candidate voxel. For voxels that cross partition boundaries, a unique gas grayscale filtering range is called according to their final partition marker to avoid the same voxel being filtered repeatedly by multiple partitions. When writing to the initial low grayscale voxel set, the spatial index, partition marker, grayscale level marker, and gated source marker of the voxel in the standardized 3D image data are saved at the same time. The gated source marker is used to record which type of partition grayscale filtering range the voxel was obtained from. After the above processing, the resulting initial low grayscale voxel set contains voxels that meet the gas grayscale conditions and retains their partition source and grayscale level information.

[0036] Step S104: Perform neighborhood grayscale detection on the initial low grayscale voxels in the initial low grayscale voxel set, construct a gas boundary shell around the initial low grayscale voxels, and mark the initial low grayscale voxels that meet the gas boundary shell existence condition and the preset voxel adjacency condition as candidate gas voxels.

[0037] In this embodiment, when performing neighborhood grayscale detection on the initial low-grayscale voxels, the partition marker, grayscale level marker, and gating source marker of each voxel in the initial low-grayscale voxel set are first read, and the shape and range of the neighborhood sampling window are determined accordingly. Specifically, for initial low-grayscale voxels originating from the cavity domain, an approximately spherical or cubic neighborhood window is used to read the grayscale changes in multiple directions on their outer periphery; for initial low-grayscale voxels originating from the pipe domain or fissure domain, a long axis neighborhood window is set along their local extension direction, and a transverse sampling area is set on both sides of the long axis; for initial low-grayscale voxels originating from the solid domain or surface domain... For low-grayscale voxels, an offset neighborhood window is set based on their adjacent boundary positions to prevent external background voxels from being mistakenly used as boundary references. After the neighborhood sampling window is determined, adjacent voxels are divided into low-grayscale voxels, transitional grayscale voxels, and non-low-grayscale voxels according to their grayscale levels within the window. Voxels with grayscale values ​​not higher than the first boundary are considered low-grayscale voxels, voxels with grayscale values ​​between the first and second boundary values ​​are considered transitional grayscale voxels, and voxels with grayscale values ​​not lower than the second boundary value are considered non-low-grayscale voxels. The first and second boundary values ​​are determined based on the boundary positions between the low-grayscale peak, the transition area, and the tissue grayscale peak in the neighborhood grayscale histogram.

[0038] Furthermore, starting from each initial low-grayscale voxel, the grayscale change sequence of neighboring voxels is read along multiple preset spatial directions. These preset spatial directions include the three-dimensional coordinate axis direction, the diagonal direction, and the main extension direction determined based on the local connectivity morphology. In specific implementation, voxel grayscale is read sequentially along each spatial direction. When the reading result shows that the voxel enters a non-low-grayscale voxel from the current low-grayscale voxel or adjacent low-grayscale voxels through one or more transitional grayscale voxels, the continuous voxel sequence in that direction is marked as a boundary transition zone. If a certain direction maintains a low-grayscale extension or directly jumps to the scan background or artifact area, that direction is not marked as a valid boundary transition zone. In the case of local noise, existing neighborhood median judgment, three-dimensional morphological opening and closing operations, or connected component filtering methods are used to remove isolated grayscale points, so that the boundary transition zone mainly reflects the continuous transition relationship between the low-grayscale area and the adjacent non-low-grayscale area. After this processing, each valid boundary transition zone is marked with a spatial direction, a grayscale level change mark, and a voxel sequence index.

[0039] Furthermore, multiple boundary transition zones obtained around the same initial low-grayscale voxel are closed-loop matched to determine the presence of a gas boundary shell. Specifically, according to the spatial orientation of the boundary transition zones, the boundary transition zones distributed around the initial low-grayscale voxel in the front, back, left, right, top, bottom, and obliquely outward directions are oriented and paired. When the boundary transition zones in multiple directions can form a continuous or nearly continuous outer envelope around the initial low-grayscale voxel, this outer envelope is considered a gas boundary shell. For tubular or fissure-shaped low-grayscale regions, the closed-loop matching does not require the formation of a completely closed structure in all directions, but rather combines its gating source marker and local extension direction to determine whether there is a boundary transition relationship around the low-grayscale region in the cross-sectional direction. For cavity-shaped or cluster-shaped low-grayscale regions, the focus is more on determining whether there are grouped boundary transition zones in multiple directions around the periphery. In this way, the low-grayscale points screened out by the grayscale threshold alone are further limited to low-grayscale regions with identifiable outer boundaries.

[0040] Finally, based on the gas boundary shell and preset voxel adjacency conditions, initial low-grayscale voxels are candidate-labeled. Specifically, for initial low-grayscale voxels that have already formed a gas boundary shell, their face adjacency, edge adjacency, or corner adjacency relationships with surrounding initial low-grayscale voxels are further read. When a voxel has a preset number or preset direction of effective adjacency with adjacent initial low-grayscale voxels, and this adjacency relationship does not cross closed boundaries, background regions, or disconnection markers, it is labeled as a candidate gas voxel. The preset number or preset direction depends on the different anatomical regions. The typical morphology of the gas region and the statistical results of the samples are determined. For initial low-grayscale voxels that have not formed a gas boundary shell, or voxels that are connected to other low-grayscale voxels only through a narrow low-grayscale channel in a single direction and whose outer boundary transition relationship is discontinuous, they are not written into the candidate gas voxel set, but into the candidate exclusion set. When writing into the candidate gas voxel set, the spatial index, partition label, gray level label, gating source label, boundary shell label and adjacency type label of the voxel are retained simultaneously to obtain candidate gas voxel data for subsequent topology graph construction.

[0041] It should be noted that the preset voxel adjacency conditions include three types of adjacency relationships: face adjacency, edge adjacency, and corner adjacency, and the judgment is made according to the principle of face adjacency first, edge adjacency second, and corner adjacency as an auxiliary.

[0042] Step S105: Establish connected edges based on the spatial adjacency relationship between the candidate gas voxels, use the disconnection marker as the blocking condition for the connected edges, and write the anatomical partition where the candidate gas voxel is located, the gas boundary shell type, the adjacency direction and the connected scale into the attribute items of the corresponding connected edge to construct a candidate gas topology graph containing candidate gas voxels, connected edges and spatial adjacency relationship.

[0043] In this embodiment, after obtaining the set of candidate gas voxels, an adjacency retrieval relationship is first established based on the spatial index of each candidate gas voxel in the standardized 3D image data. Specifically, a 3D raster index or voxel hash index is used to write the position coordinates of each candidate gas voxel into the index table, and then its surrounding adjacent voxels are read according to a preset neighborhood type. The preset neighborhood type includes face adjacency, edge adjacency, and corner adjacency, where face adjacency means that two candidate gas voxels have a common voxel face, edge adjacency means that two candidate gas voxels have a common voxel edge, and corner adjacency means that two candidate gas voxels are only connected at the corner point. Implementably, face adjacency relationship is determined first, then edge adjacency relationship, and then corner adjacency relationship. For two candidate gas voxels that meet the adjacency condition, a corresponding connected edge is established, and the two candidate gas voxels are respectively used as the starting node and ending node of the connected edge, thereby forming an initial connected graph with candidate gas voxels as nodes and spatial adjacency relationship as edges.

[0044] Furthermore, during the process of establishing connected edges, the disconnection markers formed during the aforementioned boundary continuity screening process are read, and these disconnection markers are used as blocking conditions for the generation or retention of connected edges. In specific implementation, if two candidate gas voxels satisfy the spatial adjacency, edge adjacency, or corner adjacency relationship, but there are situations where there is a discontinuity in the gas boundary shell, a one-way connection of low grayscale narrow-path channels, crossing of partition closed boundaries, crossing of background removal areas, or being pre-marked as disconnection positions, then a connected edge is not established, or the already established initial connected edge is marked as a blocking edge. For candidate gas voxels located at the boundaries of different partitions, the grayscale transition state and shell direction at the boundary position are further read. When the boundary still maintains a continuous low grayscale transition and does not trigger the disconnection condition, the cross-region connected edge is retained. When there is only point contact or the boundary shells are separated at the boundary, the cross-region connected edge is blocked. Through the above processing, the connectivity relationship is no longer determined solely by whether the voxels are adjacent, but is simultaneously constrained by boundary continuity and disconnection markers.

[0045] Furthermore, attribute items are written to the retained connected edges to construct a candidate gas topology graph. Specifically, for each connected edge, the partition markers of the candidate gas voxels at both ends, the gas boundary shell type, the adjacency direction, the adjacency type, and the connectivity scale are recorded. Among them, the partition markers of the candidate gas voxels are used to characterize that the connected edge is located in one of the cavity domain, tube domain, solid domain, surface domain, or fissure domain, or spans two of these regions. The gas boundary shell type is used to characterize that the shell corresponding to the candidate gas voxels at both ends of the connected edge is a closed shell, a semi-closed shell, an extended shell, or a boundary shell. The adjacency direction is determined by the coordinate difference between the two voxels. The connectivity scale is determined by the number of voxels, the extension length, or the local cross-sectional range within the connected domain of the connected edge. After the attribute writing is completed, the candidate gas voxels, connected edges, and connected edge attributes are uniformly stored as graph structure data to obtain the candidate gas topology graph. This candidate gas topology graph not only retains the spatial position of low-grayscale gas voxels, but also records the connection direction, connection strength, and cross-regional relationship between voxels.

[0046] like Figure 2 As shown, the construction process of the candidate gas topology map includes a transformation process from three-dimensional image voxel slicing to adjacency relation extraction, and then to topology map representation. Figure 2 The left side shows an example of a 3D image voxel slice, where dark voxels represent low-grayscale gas voxels obtained after gas grayscale screening, and light-gray voxels represent non-gas voxels. In practice, low-grayscale voxels are first extracted from the standardized 3D image data according to the gas grayscale screening range, and the coordinate position, layer number, and voxel index of each low-grayscale voxel in 3D space are retained. Since low-grayscale voxels may be distributed in clusters, bends, or discontinuously in 3D space, the connectivity relationship in the 2D slice cannot be used as the basis for judgment. Instead, their adjacent states need to be further analyzed in the 3D voxel space.

[0047] Figure 2 The middle section illustrates the process of establishing adjacency relationships for candidate gas voxels. Specifically, a neighborhood detection range is set around each low-grayscale voxel to determine whether there is a face adjacency, edge adjacency, or corner adjacency relationship between it and its surrounding low-grayscale voxels. In the figure, blue dots represent candidate gas voxels that have been identified as topological nodes, green solid lines represent face adjacency, orange dashed lines represent edge adjacency, and purple dotted lines represent corner adjacency. For a local region composed of multiple adjacent candidate gas voxels, the boundary of its gas connectivity region is marked by a dashed box. This boundary is used to define the spatial range of the same candidate gas region. If two low-grayscale voxels are close in position but do not meet the preset voxel adjacency conditions, or are merely isolated noise points, then no connectivity relationship is established.

[0048] Figure 2 The right side shows a candidate gas topology graph abstracted from candidate gas voxels and their adjacency relationships. In practice, each candidate gas voxel is converted into a topology node, such as V1 to V in the figure. 12 The spatial adjacency relationship between candidate gas voxels that meets the adjacency condition is converted into a connected edge. Different line types correspond to different adjacency types: face adjacency edge indicates that two voxels have a relatively stable spatial contact relationship, edge adjacency edge indicates that two voxels form a connection in the edge direction, and corner adjacency edge indicates that two voxels only contact in the corner direction.

[0049] Step S2: Based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency between them, and combined with the putrefaction correlation parameters, calculate the corresponding putrefaction interference potential value. According to the spatial propagation gradient of the putrefaction interference potential value in the candidate gas topology map, divide the candidate gas voxel into multiple putrefaction interference partitions, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone.

[0050] The specific steps of step S2 are as follows: Step S201: Obtain the putrefaction correlation parameters corresponding to the three-dimensional virtual anatomical image data, map the putrefaction correlation parameters to the preset anatomical standard space, and write the corresponding external putrefaction markers for the candidate gas voxels in the candidate gas topology map. The putrefaction correlation parameters include the ex vivo interval, storage temperature, storage humidity, sealing status, cooling record, and pre-collection transport status.

[0051] In this embodiment, putrefaction-related parameters are obtained from inspection records, image acquisition registration information, refrigeration or storage records, transportation records, and on-site environmental records, and are linked to the same object identifier as the three-dimensional virtual anatomical image data. Specifically, the excision interval is determined based on the temporal relationship between the excision time, discovery time, or registration time and the image acquisition time. Storage temperature and storage humidity are obtained from storage environment monitoring records or pre-acquisition environment records. The sealing status is categorized and recorded based on whether it is in a sealed bag, wrapping material, container, or open state. Cooling records include whether it is refrigerated, the start and end times of refrigeration, cooling interruption, and rewarming. Pre-acquisition transportation status includes transportation duration, transportation environment, whether it is transported multiple times, and whether temperature environment switching occurs. For parameters from different sources, they are first sorted by time sequence and missing items are marked, and then converted into a unified putrefaction-related parameter table so that each parameter can correspond to the image acquisition time, image sequence number, and the three-dimensional volume data to be processed.

[0052] Furthermore, the putrefaction correlation parameters are converted into external putrefaction markers written into the candidate gas topology map. In specific implementation, the putrefaction correlation parameters are divided into time-related parameters, environmental parameters, storage parameters, cooling parameters, and transport parameters according to their properties. Among them, time-related parameters are used to characterize the time stage from ex vivo to collection, environmental parameters are used to characterize the temperature and humidity conditions before collection, storage parameters are used to characterize the degree of gas escape restriction, cooling parameters are used to characterize the stage in which the putrefaction process is suppressed or restored, and transport parameters are used to characterize the external disturbances and environmental switching before collection. For continuously recorded temperature, humidity, or refrigeration information, it is segmented and organized according to the time period before collection. For information such as storage status and transport status, it is converted into preset category markers. After the above processing, the obtained external putrefaction markers do not directly change the image grayscale, but are used as external condition inputs when calculating the putrefaction interference potential value of candidate gas voxels.

[0053] Furthermore, the external putrefaction markers are mapped to a preset anatomical standard space and written into the candidate gas voxels in the candidate gas topology map. This can be implemented by first reading the index correspondence between the original image coordinate space and the preset anatomical standard space saved in step S102, determining the partition position, hierarchical position, and connected domain affiliation of each candidate gas voxel in the preset anatomical standard space; then, according to the applicable range of the putrefaction-related parameters, the overall parameters are written into all candidate gas voxels, and the parameters related to storage posture, local pressure, sealing contact surface, or environmental exposure direction are written into the corresponding range of candidate gas voxels according to their spatial orientation and partition position in the preset anatomical standard space. After writing, each candidate gas voxel, in addition to retaining grayscale levels, boundary shells, partition markers, and connected edge attributes, also carries a corresponding external putrefaction marker. The external putrefaction marker is used in subsequent steps to distinguish candidate gas regions with different putrefaction interference tendencies due to changes in time, temperature and humidity, sealing, cooling, or transportation conditions.

[0054] Step S202: Based on each candidate gas voxel in the candidate gas topology map, and in conjunction with the external putrefaction marker, generate a local putrefaction attribute label for that candidate gas voxel.

[0055] In this embodiment, before generating local decay attribute labels, the node information and adjacency information of candidate gas voxels are read one by one from the candidate gas topology map. Specifically, for each candidate gas voxel, its gray level label, its region label, gas boundary shell type, gating source label, number of adjacent candidate gas voxels, connected edge direction, and scale information of its connected domain are read. Among them, the gray level label is used to characterize the low gray level of the voxel, the region label is used to characterize the voxel's location in a cavity domain, tube domain, solid domain, surface domain, or fissure domain, the gas boundary shell type is used to characterize the boundary state between the low gray level region around the voxel and the adjacent non-low gray level region, and the connected edge direction and connected domain scale are used to characterize the spatial form of the voxel's distribution, whether it is clustered, chain-like, sheet-like, or isolated.

[0056] Furthermore, the external decay markers carried by each candidate gas voxel are correlated with its local topological information. Specifically, if a candidate gas voxel has multiple adjacent directions within its connected domain, and there are many similar low-grayscale voxels around it with a locally enveloping boundary shell, then the local spatial morphology of the voxel is recorded as an aggregation attribute. If the candidate gas voxel is continuously connected to multiple candidate gas voxels along the same or approximately the same direction, and the connected edges have a clear extension sequence, then its local spatial morphology is recorded as an extension attribute. If the number of effective connected edges between the candidate gas voxel and its surrounding candidate gas voxels is small, or it is isolated by blocked edges, closed boundaries, or partition boundaries, then its local spatial morphology is recorded as an isolation attribute. The external decay markers formed by the detachment interval, storage temperature, storage humidity, sealing state, cooling records, and pre-collection transport state are then combined with the above-mentioned aggregation, extension, or isolation attributes to form an attribute combination that simultaneously reflects external conditions and local image performance.

[0057] Furthermore, based on the attribute combination, local putrefaction attribute labels for candidate gas voxels are generated, and these labels are written into the corresponding nodes of the candidate gas topology graph. The implementation method is as follows: candidate gas voxels with high putrefaction correlation parameters and exhibiting clustered or sheet-like aggregation locally are marked as having an in-situ generation tendency label; candidate gas voxels with high putrefaction correlation parameters and exhibiting unidirectional or multi-segment continuous extension along connected edges are marked as having a migration and diffusion tendency label; candidate gas voxels disconnected from the main connected domain, with relatively independent boundary shells, and whose spatial distribution cannot be explained by external putrefaction labels are marked as having a retention observation tendency label. For candidate gas voxels exhibiting different spatial morphologies simultaneously within the same connected domain, labels are written separately for each voxel node, and the label change position between adjacent voxels is recorded in the connected edge attributes. The resulting local putrefaction attribute labels include both the voxel's own grayscale and boundary features, as well as its adjacent position in the candidate gas topology graph and external putrefaction conditions.

[0058] Step S203: Assign a corruption interference potential value to each candidate gas voxel according to the local corruption attribute label. Specifically, assign a first type of corruption interference potential value to candidate gas voxels located in closed cavities, fissures, or surface areas and distributed in a cluster. Assign a second type of corruption interference potential value to candidate gas voxels that extend in a chain along pipes or interlayer gaps. Assign a third type of corruption interference potential value to candidate gas voxels that are adjacent to sudden fracture boundaries, penetration paths, or local abnormal cavities. The corruption interference potential value is used to measure the strength of the corruption effect on a candidate gas voxel.

[0059] In this embodiment, the assignment of corruption interference potential is based on the node labels in the candidate gas topology graph, rather than directly determining it based solely on a single grayscale threshold. Specifically, the local corruption attribute label of each candidate gas voxel is first read, and the partition, connected domain morphology, boundary shell type, adjacency direction, number of connected edges, and external corruption markers of the voxel are read simultaneously. For the same candidate gas connected domain, the distribution morphology of each voxel within it is determined, such as clustered, chain-like, sheet-like, or isolated. Then, combined with the spatial attribution of the connected domain to the closed cavity, fissure, surface, pipe, or interlayer gap, the corruption interference category of each candidate gas voxel within the connected domain is determined. The above reading and classification process is implemented using existing three-dimensional connected domain analysis, skeletonization analysis, adjacency matrix traversal, and local morphology classification methods. Among them, three-dimensional connected domain analysis is used to identify the same gas region, skeletonization analysis is used to determine whether there is an extending principal axis, and adjacency matrix traversal is used to determine the propagation connection relationship between voxels.

[0060] Specifically, for candidate gas voxels located in closed cavities, fissures, or surface regions and distributed in clusters, a first-type corruption interference potential value is assigned to them. In practice, the candidate gas topology map is first searched for a set of nodes with multi-directional connected edges, a locally enclosed outer boundary shell, and similar low-grayscale voxels clustered within a certain spatial range. When the partition to which the node set belongs is a closed cavity, fissure, or surface region, and its external corruption markers show changes in time, temperature, humidity, sealing, or cooling conditions related to the corruption process, the candidate gas voxels in the node set are written into the first-type corruption interference potential value. This type of potential value is used to characterize the property that the gas is more likely to form in situ in a local area. For candidate gas voxels that extend in a chain-like manner along pipes or interlayer gaps, it is first identified whether their connected edges are arranged along one or more continuous directions, and it is determined whether adjacent nodes have the characteristics of progressive connected scale or consistent extension main axis. When the above conditions are met, the corresponding candidate gas voxel is assigned a second-type corruption interference potential value, which is used to characterize the property of gas migration and diffusion along existing channels or interlayer paths.

[0061] Furthermore, for candidate gas voxels adjacent to sudden fracture boundaries, penetration paths, or local anomalous cavities, a third type of putrefaction interference potential is assigned. In specific implementation, firstly, it is read whether there are discontinuous boundaries, sharp turning boundaries, cross-layer penetration boundaries, or local cavity boundaries in the boundary shell around the candidate gas voxel. Then, it is determined whether there are blocking edges, partition boundaries, or propagation break relationships between the candidate gas voxel and the first or second type of putrefaction interference potential region. When the candidate gas voxel is isolated or spatially adjacent to the anomalous boundary, and its distribution cannot be continuously explained by in-situ cluster generation or chain-like migration and diffusion relationships, it is assigned a third type of putrefaction interference potential. After the assignment is completed, each candidate gas voxel carries the corresponding putrefaction interference potential category and is stored in the candidate gas topology map together with the original partition markers, boundary shell markers, connected edge attributes, and external putrefaction markers.

[0062] Step S204: Using the connected edges in the candidate gas topology graph as the propagation path, determine the propagation order of the corruption interference potential in the candidate gas topology graph based on the corruption interference potential, adjacency direction and connectivity scale between adjacent candidate gas voxels, and form the corresponding spatial propagation gradient.

[0063] In this embodiment, before forming the spatial propagation gradient, the propagation attributes of the connected edges in the candidate gas topology graph are determined. Specifically, the corruption interference potential value category, adjacency direction, adjacency type, connectivity scale, partition label, and gas boundary shell state of the candidate gas voxels at both ends of the connected edge are read one by one. When the boundary shell between the two candidate gas voxels is continuous, the partition transition is traceable, and the corruption interference potential values ​​have the same category or adjacent category continuity, the connected edge is marked as a propagable edge. If there is a shell interruption, crosses the closed partition boundary, only contacts through corners, abrupt changes in connectivity scale, or a disconnection mark formed in the previous steps between the two candidate gas voxels, the connected edge is marked as a blocking edge. This process can be achieved by using existing graph traversal, adjacency matrix labeling, and three-dimensional connected domain analysis methods. Its purpose is to ensure that the subsequent propagation sequence unfolds only along connected edges with spatial continuity, and not to take accidental contact or cross-boundary pseudo-connectivity as the propagation path.

[0064] Furthermore, after classifying the connected edges, the propagation starting point is determined from the candidate gas topology graph. In practice, within each connected domain, a set of candidate gas voxels whose corruption interference potential values ​​are clustered in the same category is retrieved, and the number of connected edges, the distribution of adjacent directions, and the partition label of each candidate gas voxel in the set are read. For candidate gas voxels with a large number of connected edges, concentrated circumferential adjacency, and high consistency of corruption interference potential value categories, they are selected as candidates for propagation starting points. Then, combined with their respective partitions, connected domain scales, and the integrity of the gas boundary shell, the final propagation starting point is determined from the candidates. When multiple propagation starting points exist, they are first classified according to the partition label, and then the processing order of the propagation starting points is determined according to the order of connected scale from large to small, boundary shell from complete to incomplete, and the range of similar potential value clusters from concentrated to discrete, thereby avoiding confusion between multiple cluster regions in the same propagation round.

[0065] Furthermore, using the determined propagation starting point as the initial node, adjacent candidate gas voxels are read step-by-step along the propagable edges. Specifically, starting from the propagation starting point, adjacent candidate gas voxels directly connected to it and whose connected edges are marked as propagable edges are read first. Then, it is determined whether the adjacent candidate gas voxel satisfies the continuation relationship of the corruption interference potential value, the consistency of the adjacent direction, and the connection scale connection relationship with the current node. When there are multiple propagable directions, candidate gas voxels with consistent corruption interference potential value category, small deviation between the adjacent direction and the current propagation direction, and gradual change in the connection scale are selected as the next propagation node. For chain-like or branch-like connected components, breadth-first traversal or depth-first traversal in graph search is used as the basis, and the traversal order is constrained by the connected edge attributes. For cluster-like connected components, the circumferential nodes close to the propagation starting point are processed first, and then the process is expanded to the outer nodes. After each node reading is completed, the propagation source, propagation direction, and propagation order of the candidate gas voxel are written into the candidate gas topology graph.

[0066] Finally, a spatial propagation gradient is generated based on the written propagation order. In specific implementation, the propagation starting point and candidate gas voxels located at the same type of cluster core as the propagation starting point are classified as the origin layer; candidate gas voxels connected to the origin layer through a propagable edge and whose putrefaction interference potential remains continuous are classified as the transition layer; candidate gas voxels that continue to extend along the main propagation direction or branch propagation direction are classified as the extension layer; and candidate gas voxels located at the end of the propagable path and no longer connected to new propagable edges are classified as the termination layer. For candidate gas voxels that are isolated by blocked edges and cannot be reached from any propagation starting point along a propagable edge, they are not included in the above propagation layers but are classified as independent layers. After this processing, the candidate gas voxels in the candidate gas topology map not only have putrefaction interference potential categories but also have a spatial propagation gradient composed of the origin layer, transition layer, extension layer, termination layer, and independent layers.

[0067] like Figure 3 As shown, step S205: Candidate gas voxels are partitioned and labeled according to the spatial propagation gradient, including in-situ putrefaction generation region, putrefaction migration and diffusion region, and diagnostic gas retention region. Candidate gas voxels located at the cluster center and whose putrefaction interference potential values ​​are transmitted in the same way between adjacent nodes are labeled as in-situ putrefaction generation region. Candidate gas voxels extending outward from the cluster center along the spatial propagation gradient are labeled as putrefaction migration and diffusion region. Candidate gas voxels that are disconnected from the spatial propagation gradient and have a third type of putrefaction interference potential value are labeled as diagnostic gas retention region.

[0068] In this embodiment, after generating the spatial propagation gradient, the candidate gas voxels in the candidate gas topology graph are first organized node by node to form a voxel determination sequence. Specifically, the gradient level of each candidate gas voxel is read sequentially, i.e., the origin layer, transition layer, extension layer, termination layer, or independent layer. At the same time, the corruption interference potential category, gas state type, connected edge type, number of propagable edges, number of blocking edges, and gas boundary shell type corresponding to the voxel are read. For candidate gas voxels within the same connected domain, they are arranged from first to last according to the propagation order. For candidate gas voxels within different connected domains, they are arranged according to the propagation starting point order, connected domain scale, and the partition label. The above organization process is implemented by traversing existing graph data, connecting domain numbering, and node attribute sorting to obtain the voxel determination sequence.

[0069] Further, in-situ putrefaction generation regions are identified in the voxel determination sequence. In specific implementation, candidate gas voxels located in the origin layer are read first, and it is determined whether they are continuously adjacent to candidate gas voxels of the same putrefaction interference potential value category through a propagable edge. When multiple candidate gas voxels in the origin layer form a ring, cluster, island, or multi-directional aggregation distribution around the same propagation starting point, these candidate gas voxels are marked as the same in-situ putrefaction generation region. For candidate gas voxels at the edge of the origin layer, if they still maintain the same putrefaction interference potential value continuity relationship with the propagation starting point, and their boundary shell is not cut off by the blocking edge, they are incorporated into the corresponding in-situ putrefaction generation region. If they are only connected to the origin layer through corner contact or across the blocking edge, they are not incorporated into the region. After this processing, each in-situ putrefaction generation region has a corresponding propagation starting point, a set of voxels of the same potential value, and a boundary shell range.

[0070] Furthermore, the putrefaction migration and diffusion zone is identified along the spatial propagation gradient. Specifically, starting with the marked in-situ putrefaction generation zone, subsequent candidate gas voxels are read sequentially in the order of origin layer, transition layer, extension layer, and termination layer. For candidate gas voxels that are directly or indirectly connected to the in-situ putrefaction generation zone through a propagable edge, it is determined whether their putrefaction interference potential value has a continuity relationship with the voxel of the previous level, and whether their adjacent direction is consistent with or approximately consistent with the current propagation direction. When a candidate gas voxel extends step by step along a tubular path, interlayer path, branch path, or narrow band path, and there is no abrupt change in the connectivity scale that does not match the propagation process, it is marked as a putrefaction migration and diffusion zone. In the case where there are branches on the same path, each branch is written into the sub-path mark of the same putrefaction migration and diffusion zone. For subsequent voxels that are truncated by blocking edges during propagation, the marking along that direction is stopped.

[0071] Finally, diagnostic gas retention zone determination is performed on candidate gas voxels that are not marked as in-situ putrefaction generation zones or putrefaction migration and diffusion zones. Specifically, candidate gas voxels located at independent levels, as well as those located at other levels but isolated by blocking edges, closed boundaries, or partition boundaries, are read to determine whether they can be reached from any in-situ putrefaction generation zone along a propagable edge. If they cannot be reached, or if the path is interrupted by a blocking edge, it is confirmed that a propagation disconnection relationship has been formed between them and the in-situ putrefaction generation zone and the putrefaction migration and diffusion zone. The gas state type of the candidate gas voxel is further read. When it exhibits an isolated type, a local cavity adjacency type, a fracture boundary adjacency type, or a through-path adjacency type distribution, it is marked as a diagnostic gas retention zone. After partition marking is completed, each candidate gas voxel in the candidate gas topology map carries a corresponding putrefaction interference partition mark. This mark, along with the original putrefaction interference potential value, propagation level, connectivity edge attributes, and boundary shell type, is saved as input for subsequent partition differentiation correction.

[0072] like Figure 3 As shown, the process of marking the partitions for corruption interference uses the candidate gas topology graph as the processing object. Candidate gas voxels are abstracted as topological nodes, and the spatial adjacency relationships between nodes that satisfy the propagation conditions are abstracted as connected edges. The vertical arrows on the left side of the figure represent the spatial propagation gradient levels, which are sequentially included from bottom to top as the origin layer L1, transition layer L2, extension layer L3, and termination layer L4. Each node, such as a1 to a6, b1 to b7, c1 to c4, d1 to d3, and e1 to e4, represents candidate gas voxels at different locations. The black solid lines between nodes represent propagable edges, the black dashed lines represent hierarchical connection edges, and the red dashed lines with crosses represent blocking edges. Thus, the spatial connectivity of the gas candidate region can be transformed into a graph structure with propagation levels and blocking relationships.

[0073] Specifically, Figure 3The green area located in the origin layer L1 represents the origin reference region for the spread of putrefactive interference. Nodes a1 to a6 in this region are at a lower propagation level and form a local continuity relationship through propagable edges. In actual processing, these nodes can be used as the basic reference region for the propagation gradient, i.e., the healthy / non-putrefactive gas region, to distinguish whether the subsequent putrefactive interference potential value diffuses outward from a certain local cluster. The blue area located in the transition layer L2 includes nodes b1 to b7, which are connected to some nodes in the origin layer L1 through hierarchical connection edges and extend upward along the spatial propagation direction. This region corresponds to the transition part in the process of putrefactive interference potential value propagating outward from the local region. The nodes maintain a continuous relationship through propagable edges, so it can be included in the initial diffusion segment of the putrefactive migration and diffusion region.

[0074] Furthermore, the orange region located in the extension layer L3 includes nodes c1 to c4. This region continues the propagation path of the transition layer L2 and is represented by a set of nodes that further diffuse along the spatial propagation gradient. The red region located in the termination layer L4 includes nodes d1 to d3, which are located at the end of the propagation path or at a higher level. In actual labeling, nodes with similar corruption interference potentials and clustered distributions near the termination layer can be labeled as in-situ corruption generation regions based on the corruption interference potentials, adjacency directions, connectivity scales, and hierarchical order of the candidate gas voxels. Nodes that are continuously connected from the extension layer to the termination layer can be labeled as corruption migration and diffusion regions. In the figure, the orange and blue regions together represent the regions formed by the diffusion of corruption interference potentials along the propagable path, while the red region represents the set of corruption interference nodes at the propagation terminal.

[0075] Furthermore, Figure 3 The purple area on the right represents the diagnostic gas retention area, which includes nodes e1 to e4. Although spatially adjacent to the main propagation area, this area is isolated from the blue, orange, or green areas by a red blocking edge. It cannot be merged into the in-situ putrefaction generation area or putrefaction migration and diffusion area along the propagation edge. In practice, if a candidate gas voxel is located at an independent level, or if it has a blocking edge, closed boundary, or propagation break relationship with the putrefaction interference propagation path, then this type of voxel is marked as a diagnostic gas retention area, and a retention lock mark is set during subsequent calibration. Figure 3 As shown, the corruption interference partitioning is not determined solely by node color or spatial distance, but rather by a combination of spatial propagation gradient, corruption interference potential, and relationships between propagable and blocking edges. This results in partitioning with clear boundaries and propagation logic between the in-situ corruption generation zone, the corruption migration and diffusion zone, and the diagnostic gas retention zone.

[0076] Step S3: Perform differential correction on multiple corruption interference zones of the candidate gas voxels respectively.

[0077] The specific steps of step S3 are as follows: Step S301: Read the putrefaction interference partition marker, putrefaction interference potential value and spatial propagation gradient level of the candidate gas voxel, and match the candidate gas voxel to the in-situ replacement strategy, migration backoff strategy or retention lock strategy according to the putrefaction interference partition marker.

[0078] In this embodiment, before performing partitioned differential correction, the node attributes and hierarchical attributes of each candidate gas voxel are read from the candidate gas topology graph. Specifically, the read content includes the putrefaction interference partition marker, putrefaction interference potential value category, spatial propagation gradient level, connected component number, number of propagable edges, number of blocking edges, and adjacency direction with adjacent candidate gas voxels. The putrefaction interference partition marker is used to indicate that the candidate gas voxel belongs to the in-situ putrefaction generation region, putrefaction migration and diffusion region, or diagnostic gas retention region. The putrefaction interference potential value is used to characterize the putrefaction interference type corresponding to the voxel in the previous judgment. The spatial propagation gradient level is used to characterize whether the voxel is in the origin layer, transition layer, extension layer, termination layer, or independent level.

[0079] Furthermore, based on the read putrefaction interference partition markers, a correction strategy matching is performed on the candidate gas voxels. Specifically, candidate gas voxels marked as in-situ putrefaction generation zones are matched to an in-situ replacement strategy. This strategy is used to generate replacement values ​​in subsequent steps using the grayscale levels of their peripheral non-gas voxels and the reference grayscale sequence of their respective partitions. Candidate gas voxels marked as putrefaction migration and diffusion zones are matched to a migration backoff strategy. This strategy segments the gas signal extending along the propagation path according to the reverse hierarchy of the spatial propagation gradient. Candidate gas voxels marked as diagnostic gas retention zones are matched to a retention locking strategy. This strategy maintains the original low grayscale performance of this type of voxel or only writes a retention marker in subsequent corrections, without participating in in-situ replacement and migration backoff.

[0080] Furthermore, after completing strategy matching, a correction strategy tag is written for each candidate gas voxel, and a corresponding correction task sequence is established. Specifically, voxels corresponding to the in-situ replacement strategy are grouped according to their partition, corruption interference potential, and connected component number; voxels corresponding to the migration backoff strategy are arranged in the order of spatial propagation gradient level from the termination layer to the origin layer; and voxels corresponding to the retention lock strategy are registered according to independent level, blocking edge, and boundary shell type. For candidate gas voxels located at the boundary of different strategies, the strategy tags and blocking edge states of their adjacent voxels are read to determine whether the voxel should be registered separately as a boundary voxel. After the above processing, a correction task sequence containing in-situ replacement tasks, migration backoff tasks, and retention lock tasks is obtained.

[0081] Step S302: For the candidate gas voxel marked as the in-situ putrefaction generation area, read the gray level of the non-gas voxel in its peripheral vicinity, and combine it with the reference gray level sequence of the partition to which the candidate gas voxel belongs to generate a virtual filling value, and write the virtual filling value into the position of the corresponding candidate gas voxel.

[0082] In this embodiment, for candidate gas voxels that have been matched to the in-situ replacement strategy, the peripheral neighborhood range corresponding to the candidate gas voxel is first determined based on its voxel spatial index and gas boundary shell marking. Specifically, a certain number of non-gas voxels are read outward from the peripheral boundary shell of the candidate gas voxel, and voxels that still belong to the candidate gas voxel set, the candidate exclusion set, the background region, and those isolated by blocking edges are excluded. Only neighboring non-gas voxels that are spatially continuous with the candidate gas voxel and whose gray level does not belong to the gas gray level screening range are retained. For clustered in-situ putrefaction generation areas, the peripheral envelope of the cluster is first determined, and then neighboring non-gas voxels are collected along different directions of the peripheral envelope. For island-shaped or locally sheet-like in-situ putrefaction generation areas, neighboring non-gas voxels are collected according to the long axis direction and the short axis direction of its boundary shell. Through this processing, a gray level reference voxel set that has a spatial adjacency relationship with the gas region to be replaced can be obtained.

[0083] Further, the grayscale levels of the adjacent non-gaseous voxels are organized, and virtual filling values ​​are generated by combining the reference grayscale sequence of the partition to which the candidate gas voxel belongs. In specific implementation, the partition markers of the cavity domain, tube domain, solid domain, surface domain, or fissure domain to which the candidate gas voxel belongs are read first, and then the reference grayscale sequence generated in the previous step is called. Subsequently, the adjacent non-gaseous voxels are grouped according to spatial direction, distance level, and grayscale stability, and voxels that obviously come from boundary artifacts, metal artifacts, high noise points, or other low grayscale anomaly areas are removed. For areas with relatively uniform grayscale distribution on the periphery, virtual filling values ​​are generated according to the main grayscale levels of the adjacent non-gaseous voxels. For areas with directional differences in grayscale on the periphery, corresponding local virtual filling values ​​are generated along different directions, and then the written value is determined according to the spatial position of the candidate gas voxel in the in-situ putrefaction generation area. The virtual filling value is not a fixed constant, but is jointly determined by the adjacent non-gaseous grayscale and the reference grayscale sequence of the partition to which it belongs, so that the grayscale of the replaced voxel corresponds to the local grayscale environment of its partition.

[0084] Furthermore, the generated virtual filling values ​​are written to the candidate gas voxel positions marked as in-situ putrefaction generation areas. This can be implemented by writing in the order from the periphery to the interior, and from small connected domains to large connected domains within the in-situ putrefaction generation area. For candidate gas voxels near the boundary shell, virtual filling values ​​generated by neighboring non-gas voxels are preferentially used. For candidate gas voxels located inside the cluster and lacking direct neighboring non-gas voxel references, internal filling values ​​are generated based on the already written peripheral voxels and their respective partition reference grayscale sequences. During the writing process, the original spatial index, original grayscale value, partition marker, putrefaction interference potential value, and correction source marker of the voxel are retained, and the correction source marker is recorded as in-situ replacement. After completing the above processing, the candidate gas voxels within the in-situ putrefaction generation area are replaced with virtual filling values ​​that match the surrounding non-gas regions and partition grayscale levels, forming the in-situ replacement result.

[0085] Step S303: For the candidate gas voxels marked as the putrefaction migration and diffusion zone, traverse them in reverse order of spatial propagation gradient, and perform segmented backtracking processing on the gas signals in the putrefaction migration and diffusion zone according to the adjacency direction and connectivity scale between adjacent candidate gas voxels.

[0086] In this embodiment, for candidate gas voxels that have been matched to the migration backtracking strategy, their hierarchical and propagation path markers in the spatial propagation gradient are first read. Specifically, candidate gas voxels marked as putrefactive migration and diffusion zones are arranged in reverse order from the termination layer, extension layer, transition layer to the origin layer, and the propagable edges, adjacency directions, connectivity scales, and partition markers between each voxel are read simultaneously. For candidate gas voxels within the same migration path, a backtracking sequence is established based on the propagation order written in the previous steps. For migration and diffusion zones with branch paths, the termination ends of each branch are first identified, and then backtracking is performed from each termination end to the common propagation source direction. This process is implemented using reverse breadth-first traversal or reverse depth-first traversal in existing graph traversal techniques, with propagable edges as valid backtracking paths and blocking edges as backtracking stopping conditions.

[0087] Furthermore, after forming the backtracking sequence, the putrefactive migration and diffusion region is segmented according to the adjacency direction and connectivity scale between adjacent candidate gas voxels. Specifically, adjacent candidate gas voxels are read one by one along the reverse backtracking sequence. When the adjacency direction between multiple consecutive voxels is consistent or approximately consistent, and the connectivity scale is within the same extension range, they are divided into the same backtracking segment. When the adjacency direction changes, the connectivity scale shows obvious contraction or expansion, crosses the boundary of a partition, or enters a branch path from the main path, that position is used as the segment boundary. For regions extending in a long, thin chain shape, they are segmented according to the path length and cross-sectional range. For regions that diffuse in a sheet-like shape and then converge, they are segmented according to the diffusion direction and the range of the local connected domain. Each backtracking segment obtained in this way has a starting and ending voxel, backtracking direction, adjacency type, and connectivity scale label.

[0088] Furthermore, segmented backtracking processing is performed on the gas signals within each backtracking segment. This can be implemented by starting with the backtracking segment corresponding to the termination layer and processing each segment sequentially from far to near. For backtracking segments far from the origin layer and with small connectivity scales, the grayscale levels of their surrounding non-gas voxels are read, generating local backtracking values ​​and writing them to the corresponding candidate gas voxel positions. For backtracking segments located on the main extension path, backtracking values ​​are generated segment by segment, combining the processing results of their adjacent previous-level backtracking segments, their adjacency direction, and the grayscale sequence of their respective partitions. For transitional segments close to the in-situ putrefaction generation zone, the grayscale connection relationship between the segment and the in-situ replacement result is preserved, and this segment is recorded as a transitional backtracking segment. Upon completion of the writing process, the original grayscale value, backtracking segment number, backtracking direction, correction source marker, and propagation level marker are retained for each processed voxel. The correction source marker is recorded as migration backtracking. After the above processing, the gas signals extending along the propagation path within the putrefaction migration and diffusion zone are segmented backtracked according to the reverse relationship of the spatial propagation gradient, forming migration backtracking results.

[0089] Step S304: For candidate gas voxels marked as diagnostic gas retention areas, read their independent levels and blocking edge markers; when there is a blocking edge between the candidate gas voxel and the in-situ putrefaction generation area or putrefaction migration and diffusion area, set a retention lock marker for the candidate gas voxel.

[0090] In this embodiment, for candidate gas voxels matched to the retention and locking strategy, the partition marker, spatial propagation gradient level, putrefaction interference potential, gas state type, and connectivity edge attributes with adjacent candidate gas voxels are first read from the candidate gas topology map. Specifically, if a candidate gas voxel is marked as a diagnostic gas retention area in the preceding partition, its independent level is read first. For candidate gas voxels that are not in an independent level but are located at the edge of the propagation path, the presence of blocking edges between them and the in-situ putrefaction generation area and the putrefaction migration and diffusion area is read. The blocking edges may originate from discontinuous boundary shells, crossing closed boundaries, only corner contact, abrupt changes in connectivity scale, or preceding disconnection markers. Through this reading process, the actual propagation relationship between the diagnostic gas retention area and the putrefaction-related gas area is determined, rather than judging whether retention is necessary solely based on spatial distance.

[0091] Furthermore, the locking conditions for candidate gas voxels within the diagnostic gas retention area are determined. Specifically, if a candidate gas voxel is at an independent level and cannot be traced back to the in-situ putrefaction generation area or putrefaction migration and diffusion area through a propagable edge, it is determined to be a propagation disconnected voxel. If the candidate gas voxel is spatially adjacent to the in-situ putrefaction generation area or putrefaction migration and diffusion area, but the connecting edge between them is marked as a blocking edge, it is determined that there is an isolation relationship between the voxel and the putrefaction interference propagation path. For a small connected domain composed of multiple diagnostic gas retention area voxels, all connecting edges around the small connected domain are read one by one. When the connections between its periphery and the putrefaction interference zone are all blocking edges, or there is only a contact relationship that does not meet the propagation conditions, the entire small connected domain is included in the retention locking range. The above determination is implemented by reachability detection and edge attribute filtering in graph traversal, and the propagable edges, blocking edges, and propagation levels already written in the candidate gas topology graph are used as the judgment criteria.

[0092] Furthermore, for candidate gas voxels that meet the locking conditions, a retention locking mark is written. This can be implemented by ensuring that the original grayscale value of the candidate gas voxel in the original 3D virtual anatomical image data is not covered by the in-situ replacement strategy or migration backoff strategy, while writing a retention locking mark, a locking source mark, and a blocking basis mark in the corresponding node attributes. The locking source mark is used to record that the voxel originates from the diagnostic gas retention area, and the blocking basis mark is used to record that it is locked due to an independent level, a blocking edge, or a propagation break relationship. For voxels located at the edge of the retention locking area and at the boundary with other correction areas, the original grayscale is read preferentially according to the retention locking mark in the subsequent fusion step, and adjacent in-situ replacement values ​​or migration backoff values ​​are prohibited from being written to the voxel position. After the above processing, the diagnostic gas retention area forms a clear locking result, which is entered into the subsequent correction result fusion process along with the in-situ replacement result and the migration backoff result.

[0093] Step S305: At the boundary of different corruption interference zones, determine the correction assignment of the boundary voxels based on the zone marker, boundary shell type and spatial propagation gradient level, and merge the candidate gas voxels after in-situ replacement, migration back and retention locking into the zone correction result.

[0094] In this embodiment, after completing in-situ replacement, migration rollback, and retention locking processes, the partition boundaries in the candidate gas topology map are first scanned to determine the boundary positions between different putrefaction interference partitions. Specifically, adjacent voxels are read according to the spatial index of the candidate gas voxels. When adjacent voxels carry different markers from in-situ putrefaction generation zone markers, putrefaction migration and diffusion zone markers, or diagnostic gas retention zone markers, the adjacent positions are registered as partition boundary positions. For candidate gas voxels located on both sides of the boundary position, their boundary shell type, connectivity edge type, spatial propagation gradient level, and whether there is a blocking edge are further read. If the boundary voxel is located between the in-situ putrefaction generation zone and the putrefaction migration and diffusion zone, and the two are continuously connected by a propagable edge, the boundary position is registered as a propagation continuous boundary. If the boundary voxel is located between the diagnostic gas retention zone and any putrefaction interference correction zone, and there is a blocking edge or propagation disconnection relationship between the two, the boundary position is registered as a retention isolation boundary. Through this process, clear partition boundary data can be provided for the subsequent assignment of boundary voxels.

[0095] Furthermore, the correction assignment of boundary voxels is determined based on the partition marker, boundary shell type, and spatial propagation gradient level. In specific implementation, for boundary voxels located at the edge of the in-situ putrefaction generation zone and still at the origin layer or transition start position, if their boundary shell is continuous with the outer envelope of the in-situ putrefaction generation zone, they are assigned to the in-situ replacement result, and the virtual fill value generated by the in-situ replacement step is called. For boundary voxels located at the edge of the putrefaction migration and diffusion zone and extending from the transition layer to the termination layer along the spatial propagation gradient, if their connected edge is consistent with the direction of the migration path, they are assigned to the migration backtracking result, and the backtracking value of the corresponding backtracking segment is called. For boundary voxels with a retention lock marker, even if they are spatially adjacent to the in-situ putrefaction generation zone or the putrefaction migration and diffusion zone, as long as there is a blocking edge, closed boundary, or propagation disconnection relationship between the two, they are assigned to the retention lock result, and their original grayscale value and lock marker are retained. In the case of multiple optional assignments for the same boundary position, the unique correction assignment is determined according to the order of retention lock priority, blocking edge priority, consistent propagation level priority, and continuous boundary shell priority.

[0096] Furthermore, the candidate gas voxels after in-situ replacement, migration rollback, and retention locking are merged into a partitioned correction result. This can be achieved by establishing a correction result table corresponding to the spatial index of the candidate gas voxels, writing the virtual fill value obtained from in-situ replacement, the segmented rollback value obtained from migration rollback, and the original grayscale value obtained from retention locking to the corresponding voxel positions. For the aforementioned boundary voxels, the corresponding correction value and correction source marker are written according to the determined correction attribution. During the writing process, the original grayscale value, corrected grayscale value, putrefaction interference partition marker, boundary shell type, spatial propagation gradient level, connected edge state, and correction source marker of each voxel are simultaneously saved. After merging, a partitioned correction result is formed. This result includes both the replaced in-situ putrefaction generation voxels, the rolled-back putrefaction migration and diffusion voxels, and the locked-and-retained diagnostic gas retention voxels.

[0097] Step S4: The partition correction results after partition differentiation correction are fused with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

[0098] The specific steps of step S4 are as follows: Step S401: Read the candidate gas voxels in the partition correction results after in-situ replacement, migration rollback and retention locking, and according to the position of the candidate gas voxels in the preset anatomical standard space, map them to the original voxel positions in the original three-dimensional virtual anatomical image data to generate a set of voxels to be fused and corrected.

[0099] In this embodiment, after the partition correction results are formed, the corrected grayscale value, correction source marker, putrefaction interference partition marker, spatial propagation gradient level, and voxel position in the preset anatomical standard space of each candidate gas voxel are read first. Specifically, the correction source marker includes in-situ replacement, migration rollback, and retention lock. The voxel corresponding to in-situ replacement reads its virtual fill value, the voxel corresponding to migration rollback reads its segmented rollback value, and the voxel corresponding to retention lock reads its original grayscale value and lock marker. During reading, a correction voxel list is established according to the spatial index of the candidate gas voxels, and voxels that are not written into the correction results, voxels in the candidate exclusion set, and non-gas voxels that are only used as neighborhood references and do not participate in the correction are excluded, thereby obtaining a record containing only the voxels that actually need to participate in the fusion.

[0100] Furthermore, based on the coordinate mapping relationship saved during the previous registration process, the candidate gas voxels are mapped back to the original three-dimensional virtual anatomical image data from the preset anatomical standard space. In specific implementation, the transformation parameters and index correspondence table between the original image coordinate space and the preset anatomical standard space established in step S102 are read. For candidate gas voxels that correspond one-to-one with the original voxels during the registration and resampling process, they are directly located to the original voxel position according to the index correspondence relationship. For cases where one standard space voxel corresponds to multiple original voxels or multiple standard space voxels correspond to one original voxel due to scale uniformity or non-rigid registration, the original voxel position is determined based on the voxel center distance, voxel coverage ratio, or nearest neighbor mapping relationship. During the mapping process, the standard space position, original space position, and correction source mark of each voxel are retained.

[0101] Furthermore, the correction voxel records corresponding to the original locations are summarized to generate a set of correction voxels to be fused. One possible approach is to use the voxel index of the original 3D virtual anatomical image data as the key, and write the corresponding corrected grayscale value, original grayscale value, correction source marker, putrefaction interference partition marker, spatial propagation gradient level, and retention / lock marker into the same record item. When multiple correction voxels are mapped to the same original voxel location, the final record is determined according to the rule of priority for retention / lock, priority for boundary assignment, second priority for migration / retrogression, and third priority for in-situ replacement. Other mapping records are retained as auxiliary tracing information. After summarization, a set of correction voxels to be fused is obtained, where each record can be located from the original 3D virtual anatomical image data to its corresponding voxel.

[0102] Step S402: Using the original three-dimensional virtual anatomical image data as a base, retain the original voxel grayscale that has not been written into the set of voxels to be fused for correction. For the original voxels that have been written into the set of voxels to be fused for correction, call the in-situ replacement value, migration rollback value or retention lock value according to their correction source mark to form layered write-back image data.

[0103] In this embodiment, the original three-dimensional virtual anatomical image data is used as the write-back base. First, the voxel matrix, spatial orientation, slice thickness, pixel spacing, gray level depth, and image sequence identifier of the original image are copied to form the write-back base data with the same size and coordinates as the original image. Specifically, the voxel index of the original three-dimensional virtual anatomical image data is read voxel by voxel, and it is determined whether the voxel index exists in the voxel set to be fused and corrected. For the original voxels that are not written into the voxel set to be fused and corrected, their original gray level value, spatial index, and image metadata remain unchanged and they do not participate in the subsequent gray level replacement.

[0104] Furthermore, for the original voxels written into the set of correction voxels to be fused, the corresponding correction value is called according to the correction source marker in the voxel record. In specific implementation, when the correction source marker is in-situ replacement, the virtual fill value generated in the in-situ replacement step of the voxel is read and written to the original voxel position; when the correction source marker is migration rollback, the migration rollback value corresponding to the rollback segment to which the voxel belongs is read and written to the original voxel position; when the correction source marker is retention lock, the original grayscale value or retention lock value of the voxel is read and the low grayscale performance of the voxel in the write-back basic data is maintained, and the retention lock marker is written. In the case where there are multiple optional correction records at the same original voxel position, the final write record determined in the previous set of correction voxels to be fused is called to ensure that only one grayscale value corresponding to the correction source is written to the same voxel.

[0105] Furthermore, the data after grayscale retrieval and voxel write-back is organized into layered write-back image data. One possible approach is to retain an original grayscale layer in the write-back base data to record the original voxel grayscale before write-back; establish a correction grayscale layer to record in-situ replacement values, migration rollback values, or retention lock values; and establish a marker layer to record the correction source marker, putrefaction interference partition marker, spatial propagation gradient level, and retention lock marker for each write-back voxel. For voxels not written back, their correction grayscale layer can remain consistent with the original grayscale layer, and the marker layer can be filled with uncorrected markers. After completing the above processing, layered write-back image data is formed. This data maintains the same spatial structure as the original three-dimensional virtual anatomical image data and can distinguish the write-back results of voxels from different correction sources.

[0106] Step S403: Read the boundary position between the corrected voxels and the uncorrected voxels in the layered write-back image data, and determine the gray-level connection order of the boundary voxels according to the partition mark, boundary shell type and spatial propagation gradient level corresponding to the boundary voxels. Perform continuous reconstruction of the boundary position according to the gray-level connection order to generate gas-corrected fused image data.

[0107] In this embodiment, after obtaining the layered write-back image data, the boundary position between the corrected voxels and the uncorrected voxels is first read according to the marker layer. Specifically, the marker layer of the layered write-back image data is scanned voxel by voxel. When a voxel carries an in-situ replacement marker, a migration back marker, or a retention lock marker, and there are uncorrected marked voxels within its face adjacency, edge adjacency, or corner adjacency range, the position is registered as a boundary position. For multiple consecutive boundary voxels, they are merged into a boundary zone through three-dimensional connected component analysis. When registering the boundary position, the decay interference partition marker, boundary shell type, spatial propagation gradient level, correction source marker, and the original gray level of the adjacent uncorrected voxels of the boundary voxel are read simultaneously to form a boundary voxel record.

[0108] Furthermore, the grayscale transition sequence at the boundary is determined based on the partition markers, boundary shell types, and spatial propagation gradient levels corresponding to the boundary voxels. Specifically, for the boundary between the in-situ replacement region and the uncorrected region, the grayscale transition sequence is determined according to the direction of the uncorrected voxels, boundary shell voxels, and in-situ replacement voxels, with priority given to the grayscale levels of adjacent non-gaseous voxels. For the boundary between the migration retreat region and the uncorrected region, the transition sequence is determined according to the reverse retreat direction of the spatial propagation gradient, i.e., gradually transitioning from the termination layer or extension layer to the adjacent uncorrected region. For the boundary between the retained-locked region and the uncorrected region, the original grayscale of the retained-locked voxel is used as the anchor value, and a grayscale transition relationship is established only at the unlocked boundary voxels on its periphery, without changing the grayscale of voxels with retained-locked markers. If the same boundary zone is adjacent to multiple correction sources, it is first determined whether cross-source transition is allowed based on the blocking edge and boundary shell type. For boundary locations with blocking edges, grayscale continuation is not performed along the blocking direction.

[0109] Furthermore, the boundary positions are continuously reconstructed according to the determined grayscale connection order to generate gas-corrected fused image data. The feasible method is to sequentially read the corrected grayscale layer and the original grayscale layer along the determined connection direction within the boundary zone, connecting the grayscale of the boundary voxels layer by layer from the corrected area to the uncorrected area. For in-situ replacement zone boundaries, the hierarchical transition result between the grayscale of adjacent non-gas voxels and the virtual fill value is written into the boundary voxel. For migration backbound zone boundaries, segmented connection results are generated according to the backbound segment number and the grayscale of adjacent uncorrected voxels. For locked zone boundaries, the original grayscale of the locked voxels is maintained, and connection writing is only performed on the unlocked outer boundary voxels. After reconstruction, the reconstructed corrected grayscale layer is merged with the uncorrected original grayscale layer, and the boundary reconstruction marker, correction source marker, and locked-area marker are retained, thereby generating gas-corrected fused image data. This gas-corrected fused image data is consistent with the original three-dimensional virtual anatomical image data in terms of spatial coordinates, voxel scale, and sequence metadata.

[0110] Step S404: Remap the gas-corrected fused image data to the original image coordinate space, and simultaneously write the partition marker, correction source marker and retention lock marker of each candidate gas voxel to generate a virtual anatomical gas image after putrefaction interference correction.

[0111] In this embodiment, after generating gas-corrected fused image data, the spatial transformation relationship, resampling parameters, and voxel index correspondence table saved during the previous registration process are read first to determine the inverse mapping relationship between the gas-corrected fused image data and the original image coordinate space. Specifically, if the gas-corrected fused image data is still in the preset anatomical standard space, the inverse transformation of the rigid registration parameters, scale correction parameters, and local non-rigid correction parameters recorded in step S102 is used to map each voxel back to the coordinate position of the original three-dimensional virtual anatomical image data. If the original index has been partially written back during the previous fusion process, the voxel center coordinates, layer number, row and column number, and voxel spacing are further checked to ensure that the corrected voxel grayscale is consistent with the actual voxel position in the original image sequence. For the case where one standard space voxel corresponds to multiple original voxels, the writing is performed according to the voxel coverage range. For the case where multiple standard space voxels correspond to the same original voxel, the final writing value is determined according to the correction source marker and boundary assignment record.

[0112] Furthermore, while completing the coordinate remapping, the partition marker, correction source marker, and retention lock marker corresponding to the candidate gas voxel are written into the auxiliary marker layer or metadata table of the output image. In specific implementation, the partition marker is used to record whether the candidate gas voxel belongs to the in-situ putrefaction generation area, putrefaction migration and diffusion area, or diagnostic gas retention area in the preceding partition; the correction source marker is used to record whether the voxel is ultimately replaced by an in-situ replacement value, migration rollback value, retention lock value, or boundary reconstruction value; the retention lock marker is used to record whether the voxel is prohibited from participating in replacement or rollback processing. For original voxels that have not participated in correction, an uncorrected marker is written or no partition attribute is written; for boundary voxels, their boundary reconstruction marker and the source of their adjacent partitions are recorded at the same time.

[0113] Furthermore, the image data after backmapping and label writing is organized into a virtual anatomical gas image corrected for putrefaction interference. This can be achieved by preserving the sequence orientation, slice thickness, pixel spacing, grayscale depth, and object identifiers of the original 3D virtual anatomical image data, and outputting the corrected grayscale volume data as the main image. Simultaneously, the partition label layer, correction source label layer, retention lock label layer, and necessary coordinate mapping records are saved as supplementary data. The output format can be a medical image sequence format compatible with the original image, or it can be saved as 3D volume data and its accompanying label file. The resulting virtual anatomical gas image corrected for putrefaction interference contains both the corrected gas grayscale representation and retains the partition source and correction path of each candidate gas voxel.

[0114] Example 2: Please see Figure 4Another embodiment of the present invention provides a putrefaction interference zoning correction system for virtual anatomical gas images, comprising: a topology map construction module, a zoning module, a correction module, and a fusion module; The topology graph construction module is used to acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology graph containing candidate gas voxels, connected edges, and spatial adjacency relationships. The partitioning module is used to calculate the corresponding putrefaction interference potential value based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency relationship between them, and in combination with putrefaction correlation parameters. Based on the spatial propagation gradient of the putrefaction interference potential value in the candidate gas topology map, the candidate gas voxel is divided into multiple putrefaction interference partitions, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone. The correction module is used to perform differential correction on multiple corruption interference zones of the candidate gas voxels respectively. The fusion module is used to fuse the partition correction results after partition differentiation correction with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

[0115] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0116] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for zoning correction of putrefaction interference in virtual anatomical gas images, characterized in that, include: Acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology map containing candidate gas voxels, connected edges, and spatial adjacency relationships; Based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency between them, and combined with the putrefaction correlation parameters, the corresponding putrefaction interference potential is calculated. According to the spatial propagation gradient of the putrefaction interference potential in the candidate gas topology map, the candidate gas voxel is divided into multiple putrefaction interference zones, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone. Differential corrections were performed on multiple corruption interference zones of the candidate gas voxels; The partition correction results after partition differentiation correction are fused with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

2. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 1, characterized in that, The construction of the candidate gas topology graph, which includes candidate gas voxels, connected edges, and spatial adjacency relationships, includes: Acquire three-dimensional virtual anatomical image data, and perform voxel scale unification, gray level normalization and background voxel removal processing on the three-dimensional virtual anatomical image data to generate three-dimensional image data to be registered. From the three-dimensional image data to be registered, extract bony anatomical anchor points, body surface contour anchor points, and cavity contour anchor points. Determine the overall posture and shape boundary based on the bony anatomical anchor points, and map the three-dimensional image data to be registered to a preset anatomical standard space to obtain standardized three-dimensional image data. Based on the standardized three-dimensional image data, multiple anatomical regions are determined, and corresponding gas grayscale filtering ranges are configured for different anatomical regions. Voxels in each anatomical region are gated according to the gas grayscale filtering range to generate an initial low grayscale voxel set. Neighborhood grayscale detection is performed on the initial low grayscale voxels in the initial low grayscale voxel set to construct a gas boundary shell around the initial low grayscale voxels. The initial low grayscale voxels that meet the gas boundary shell existence condition and the preset voxel adjacency condition are marked as candidate gas voxels. Based on the spatial adjacency relationship between the candidate gas voxels, connected edges are established. The anatomical region where the candidate gas voxel is located, the gas boundary shell type, the adjacency direction and the connectivity scale are written into the attribute items of the corresponding connected edges, and a candidate gas topology graph containing candidate gas voxels, connected edges and spatial adjacency relationships is constructed.

3. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 2, characterized in that, Based on the standardized 3D image data, multiple anatomical regions are determined, and corresponding gas grayscale filtering ranges are configured for different anatomical regions. Voxels within each anatomical region are then gated according to the gas grayscale filtering ranges to generate an initial set of low-grayscale voxels, including: According to the spatial coordinates, density levels and closed contours in the preset anatomical standard space, the standardized three-dimensional image data is divided into cavity domain, canal domain, solid domain, surface domain and fissure domain, and corresponding partition markers are written for each voxel. Within each region corresponding to a partition marker, different voxels are selected to form a partition grayscale sample set, and a grayscale level sequence corresponding to the partition is generated based on the partition grayscale sample set. Based on the grayscale level sequence, spatial enclosure degree and neighborhood density difference of each partition, the corresponding gas grayscale screening range is configured respectively. Specifically, the first grayscale screening range is configured for the cavity domain, the second grayscale screening range is configured for the pipe domain, the third grayscale screening range is configured for the solid domain, and the fourth grayscale screening range is configured for the surface domain and the fissure domain. According to the partition label of each voxel, the corresponding gas grayscale filtering range is called, and the voxels in each partition are gated and filtered; the voxels that fall into the corresponding gas grayscale filtering range are written into the initial low grayscale voxel set.

4. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 3, characterized in that, Neighborhood grayscale detection is performed on the initial low-grayscale voxels in the initial low-grayscale voxel set to construct a gas boundary shell around the initial low-grayscale voxels. Initial low-grayscale voxels that satisfy the gas boundary shell existence condition and the preset voxel adjacency condition are marked as candidate gas voxels, including: Based on the partition label, gray level label, and gate source label of each initial low gray voxel in the initial low gray voxel set, a corresponding neighborhood sampling window is set, and low gray voxels, transition gray voxels, and non-low gray voxels adjacent to the initial low gray voxel are read within the neighborhood sampling window. Starting with an initial low-grayscale voxel, the grayscale change sequence of neighboring voxels is read along multiple preset spatial directions, and the voxel sequence that continuously transitions from low-grayscale voxels to non-low-grayscale voxels is marked as a boundary transition zone. Multiple boundary transition zones surrounding the same initial low-grayscale voxel are closed and matched according to the spatial direction. When multiple boundary transition zones form an outer envelope that satisfies the preset closure condition on the outer periphery of the initial low-grayscale voxel, the outer envelope is determined as the gas boundary shell corresponding to the initial low-grayscale voxel. Initial low-grayscale voxels with gas boundary shells and satisfying preset voxel adjacency conditions are marked as candidate gas voxels.

5. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 1, characterized in that, Based on each candidate gas voxel in the candidate gas topology map and their spatial adjacency relationships, and combined with the putrefaction correlation parameters, the corresponding putrefaction interference potential is calculated. According to the spatial propagation gradient of the putrefaction interference potential in the candidate gas topology map, the candidate gas voxels are divided into multiple putrefaction interference partitions, including: Obtain the putrefaction correlation parameters corresponding to the three-dimensional virtual anatomical image data, map the putrefaction correlation parameters to the preset anatomical standard space, and write the corresponding external putrefaction markers for the candidate gas voxels in the candidate gas topology map; Based on each candidate gas voxel in the candidate gas topology map, and in conjunction with the external putrefaction marker, a local putrefaction attribute label for the candidate gas voxel is generated. According to the local corruption attribute labels, each candidate gas voxel is assigned a corruption interference potential value. Using the connected edges in the candidate gas topology graph as the propagation path, the propagation order of the corruption interference potential in the candidate gas topology graph is determined based on the corruption interference potential, adjacency direction and connectivity scale between adjacent candidate gas voxels, and the corresponding spatial propagation gradient is formed. Candidate gas voxels are partitioned and labeled according to spatial propagation gradients, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone, and diagnostic gas retention zone.

6. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 5, characterized in that, The process involves using connected edges in the candidate gas topology graph as propagation paths, determining the propagation order of the corruption interference potential in the candidate gas topology graph based on the corruption interference potential, adjacency direction, and connectivity scale between adjacent candidate gas voxels, and forming the corresponding spatial propagation gradient, including: Read the corruption interference potential, adjacency direction and connectivity scale of the candidate gas voxels at both ends of each connected edge in the candidate gas topology graph, and mark the connected edges as propagable edges or blocking edges according to the continuity of the gas boundary shell and the cross-region situation. In the candidate gas topology graph, candidate gas voxels whose corruption interference potential values ​​are clustered in the same type and whose number of connected edges meets the preset clustering conditions are determined as the propagation starting points. When there are multiple propagation starting points, the order of propagation starting points is determined according to the partition, connectivity scale and gas boundary shell state. Taking the propagation starting point as the initial node, read the adjacent candidate gas voxels along the propagable edge, and determine the propagation order of the corruption interference potential value based on the continuity relationship of the corruption interference potential value, the consistency of the adjacent direction, and the connection relationship of the connectivity scale between the adjacent candidate gas voxels. According to the propagation order, candidate gas voxels are divided into origin layer, transition layer, extension layer and termination layer, and candidate gas voxels isolated by blocking edges are divided into independent levels. The origin layer, transition layer, extension layer, termination layer and independent levels form a corresponding spatial propagation gradient.

7. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 6, characterized in that, The step of partitioning and labeling candidate gas voxels based on spatial propagation gradients includes: Based on the spatial propagation gradient and the corruption interference potential of candidate gas voxels, the connected edges and the gas boundary shell type, a voxel determination sequence is generated. In the voxel determination sequence, candidate gas voxels located in the origin layer and continuously adjacent to voxels of the same type of putrefaction interference potential are marked as in-situ putrefaction generation regions. Along the spatial propagation gradient from the origin layer to the transition layer, extension layer and termination layer, candidate gas voxels connected to the in-situ putrefaction generation zone through a propagable edge are read sequentially. Candidate gas voxels whose putrefaction interference potential has a continuous relationship and whose adjacent direction is consistent with the propagation direction are marked as putrefaction migration and diffusion zones. Candidate gas voxels located at independent levels or isolated by blocked edges are read, and it is determined whether they form a propagation disconnect relationship with the in-situ putrefaction generation zone and putrefaction migration and diffusion zone. Candidate gas voxels that form a propagation disconnect relationship and have an isolated gas state type are marked as diagnostic gas retention zones.

8. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 1, characterized in that, The differential correction of multiple corruption interference regions of candidate gas voxels includes: Read the putrefaction interference partition marker, putrefaction interference potential and spatial propagation gradient level of the candidate gas voxel, and match the candidate gas voxel to the in-situ replacement strategy, migration backoff strategy or retention lock strategy according to the putrefaction interference partition marker. For candidate gas voxels marked as in-situ putrefaction generation zones, the grayscale level of the non-gas voxels around them is read, and combined with the reference grayscale sequence of the partition to which the candidate gas voxel belongs, a virtual filling value is generated and written into the corresponding candidate gas voxel position. For candidate gas voxels marked as putrefaction migration and diffusion zones, traversal is performed in reverse order of spatial propagation gradient, and the gas signals in the putrefaction migration and diffusion zones are segmented and backtracked based on the adjacency direction and connectivity scale between adjacent candidate gas voxels. For candidate gas voxels marked as diagnostic gas retention areas, their independent levels and blocking edge markers are read; when there is a blocking edge between the candidate gas voxel and the in-situ putrefaction generation area or putrefaction migration and diffusion area, a retention lock marker is set for the candidate gas voxel. At the boundary of different corruption interference zones, the correction assignment of boundary voxels is determined based on the zone marker, boundary shell type, and spatial propagation gradient level. Candidate gas voxels that have undergone in-situ replacement, migration back and retention locking are then merged into the zone correction results.

9. The method for zoning correction of putrefaction interference in virtual anatomical gas images as described in claim 1, characterized in that, The process of fusing the partition correction results (after partition differentiation correction) with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction includes: Read the candidate gas voxels in the partition correction results after in-situ replacement, migration rollback and retention locking, and according to the position of the candidate gas voxels in the preset anatomical standard space, map them to the original voxel positions in the original three-dimensional virtual anatomical image data to generate a set of voxels to be fused and corrected. Based on the original three-dimensional virtual anatomical image data, the original voxel grayscale that was not written into the set of voxels to be fused and corrected is retained. For the original voxels that were written into the set of voxels to be fused and corrected, the in-situ replacement value, migration rollback value or retention lock value is called according to their correction source mark to form layered write-back image data. Read the boundary position between the corrected voxels and the uncorrected voxels in the layered write-back image data, and determine the gray-level connection order of the boundary voxels according to the partition marker, boundary shell type and spatial propagation gradient level corresponding to the boundary voxels. Then, continuously reconstruct the boundary position according to the gray-level connection order to generate gas-corrected fused image data. The gas-corrected fused image data is remapped to the original image coordinate space, and the partition markers, correction source markers, and retention lock markers of each candidate gas voxel are written simultaneously to generate a virtual anatomical gas image after putrefaction interference correction.

10. A putrefaction interference zoning correction system for virtual anatomical gas images, used to implement the putrefaction interference zoning correction method for virtual anatomical gas images as described in any one of claims 1-9, characterized in that, include: Topology map construction module, partitioning module, calibration module, and fusion module; The topology graph construction module is used to acquire three-dimensional virtual anatomical image data, register the three-dimensional virtual anatomical image data to a preset anatomical standard space, and construct a candidate gas topology graph containing candidate gas voxels, connected edges, and spatial adjacency relationships. The partitioning module is used to calculate the corresponding putrefaction interference potential value based on each candidate gas voxel in the candidate gas topology map and the spatial adjacency relationship between them, and in combination with putrefaction correlation parameters. Based on the spatial propagation gradient of the putrefaction interference potential value in the candidate gas topology map, the candidate gas voxel is divided into multiple putrefaction interference partitions, including in-situ putrefaction generation zone, putrefaction migration and diffusion zone and diagnostic gas retention zone. The correction module is used to perform differential correction on multiple corruption interference zones of the candidate gas voxels respectively. The fusion module is used to fuse the partition correction results after partition differentiation correction with the original three-dimensional virtual anatomical image data to generate a virtual anatomical gas image after putrefaction interference correction.

Images