Multi-modal medical image fusion processing method, system, medium and equipment

Abstract

The invention discloses a multi-modal medical image fusion processing method and system, a medium and equipment, and relates to the technical field of medical image processing, and the method comprises the steps: carrying out the lesion heterogeneity partitioning and pathological coverage evaluation according to a cross-modal evidence topological chain and a mapping credible mark, and carrying out the representative deviation correction, generating an uncovered area mark and a representative correction evidence set; anti-fact evidence verification is carried out on the representative correction evidence set and the uncovered area marks, state judgment is carried out in combination with the mapping credible marks, and an evidence state diagram and a conflict handling instruction are output; and performing evidence chain constraint fusion according to the evidence state diagram and the conflict processing instruction, generating an evidence fusion labeling image, giving a re-evidence collection prompt for a conflict area, and updating the evidence state diagram and the conflict processing instruction in combination with an uncovered area label. According to the method, representative deviation correction is implemented and effective correction evidences are recombined, so that the disposal pertinence and evidence closed-loop performance of evidence fusion annotation images are improved.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a multimodal medical image fusion processing method, system, medium, and device. Background Technology

[0002] Multimodal medical image fusion processing technology is geared towards collaborative scenarios of image diagnosis and pathological review. Conventional methods typically revolve around tomographic image registration, pathological full-view image localization, lesion region extraction, and fusion annotation. By establishing cross-carrier correspondence through case identification, time records, and spatial coordinate information, and combining regional segmentation and coverage statistics, it enables lesion observation, pathological mapping, and visualization, providing structured evidence support for the joint analysis of digital pathology and medical images.

[0003] However, conventional methods still have room for improvement: on the one hand, imaging lesion areas and pathological field blocks often rely on single pairing or static mapping, lacking continuous verification along the sampling number chain, slice direction relationship and spatial adjacency relationship, and the reliability of cross-modal correspondence is easily affected by local mismatch; on the other hand, zonal pathological coverage assessment mostly adopts overall statistics or coarse-grained regional statistics, which makes it difficult to identify uncovered areas and link conflict handling at the granularity of lesion heterogeneity zonal division. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multimodal medical image fusion processing method to solve the problems that cross-modal correspondence reliability is easily affected by local mismatches and that there is insufficient linkage between regional pathological coverage assessment and conflict handling.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a multimodal medical image fusion processing method, which includes collecting multimodal medical image data of cases and pathological sampling records and establishing a correlation relationship to obtain standardized case evidence units and sampling record correlation tables; Based on standardized case evidence units and tissue sampling record association tables, a correspondence chain between imaging lesion areas and pathological field blocks is established and its credibility is verified, generating cross-modal evidence topology chains and mapping credibility markers. Based on the cross-modal evidence topology chain and mapping credibility markers, lesion heterogeneity partitioning and pathological coverage assessment are performed, and representativeness bias correction is implemented to generate uncovered area markers and representativeness correction evidence sets. Counterfactual evidence verification is performed on the representative corrected evidence set and the uncovered area markers, and the status is adjudicated in combination with the mapped credible markers, outputting the evidence status diagram and conflict resolution instructions; Based on the evidence status diagram and conflict resolution instructions, the evidence chain constraint fusion is performed to generate evidence fusion labeled images, prompts for re-examination of conflict areas are given, and the evidence status diagram and conflict resolution instructions are updated in combination with the marking of uncovered areas.

[0007] As a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the multimodal medical image data of the case includes case identifier, multimodal tomographic image, pathological full-view image, acquisition timestamp, image spatial orientation parameter, image spatial resolution parameter and pathological full-view image coordinate information; The pathological sampling record includes case identifier, sampling number, paraffin block number, slide number, slide orientation mark, sampling site and layer information.

[0008] In a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the steps for obtaining the standardized case evidence unit and the tissue sampling record association table are as follows: The multimodal medical imaging data and pathological sampling records of cases were processed by time information correction, name field unification and spatial benchmark standardization to form a standardized case dataset; Based on the standardized case dataset, a matching index is established between the sampling number and the multimodal medical image data of the case. The multimodal medical image data of the case and the pathological sampling records are associated, encapsulated, and written into the association field to generate a standardized case evidence unit and a sampling record association table.

[0009] As a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the steps of generating cross-modal evidence topology chains and mapping trustworthy markers are as follows: The location information of the imaging lesion area and the coordinate information of the pathological field block are extracted from the standardized case evidence unit, and the candidate pairing relationship between the imaging lesion area and the pathological field block is established by combining the sampling record association table to form a candidate dataset of lesion field. Based on the relationship between the lesion field candidate dataset and the tissue sampling record association table, a hierarchical candidate correspondence relationship between the imaging lesion region and the pathological field block is established, and a candidate correspondence chain set is generated. Perform number continuity verification, slice direction consistency verification, and spatial adjacency verification on the candidate corresponding chain set to form a set of credible verification chains; Based on the credibility verification chain set, the topological connection fields are organized according to the hierarchical connection relationship to generate a cross-modal evidence topological chain. Based on the verification conclusions of each candidate corresponding chain in the credibility verification chain set, a mapping credibility tag is generated.

[0010] As a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the steps of generating the uncovered area markers and representative correction evidence set are as follows: Based on the cross-modal evidence topology chain and mapping credibility markers, credibility constraint analysis is performed on the corresponding nodes of the imaging lesion region and the pathological field of view block, and the spatial coverage is organized to form a lesion coverage analysis dataset. The image lesion regions in the lesion coverage analysis dataset are partitioned according to regional difference characteristics, and partition boundaries and partition identifiers are established to generate a lesion heterogeneity partition set. Based on the lesion heterogeneity partition set and lesion coverage analysis dataset, the coverage of each partition by the pathological field block is statistically analyzed, and pathological coverage is evaluated by combining the mapping confidence label, generating the label of the uncovered area; Representative bias correction is performed based on the mapping credibility markers, the heterogeneous lesion partition set, and the uncovered area markers, and the effective correction evidence is reorganized to generate a representative correction evidence set.

[0011] In a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the steps of outputting the evidence status diagram and conflict resolution instructions are as follows: Based on the representative correction evidence set and the uncovered area markers, local perturbations that preserve structural boundaries are performed on the imaging lesion areas and pathological field blocks, and the changes in judgment before and after the perturbation are recorded to generate a counterfactual verification record set. Based on the counterfactual verification record set, the consistency and conflict of evidence in each area of ​​evidence to be verified are determined, and the missing status information is supplemented by the marking of uncovered areas to generate a set of evidence credibility conclusions. Based on the mapping of credible tags and the set of evidence credibility conclusions, the execution status adjudication and disposal rules are matched, and the evidence status diagram and conflict disposal instructions are output.

[0012] As a preferred embodiment of the multimodal medical image fusion processing method of the present invention, the steps of performing evidence chain constraint fusion based on the evidence status map and conflict resolution instructions to generate evidence fusion annotated images, providing re-evidence prompts for conflict areas, and updating the evidence status map and conflict resolution instructions in conjunction with the marking of uncovered areas are as follows: Based on the evidence status diagram and conflict resolution instructions, the status category and resolution requirements of each area are read, and the corresponding evidence chain constraint rules are matched according to the status category to form an evidence chain constraint fusion task set; Based on the evidence chain constraint fusion task set, the case multimodal medical image data is called, and evidence chain constraint fusion processing is performed on each image lesion area to form an intermediate set of regional fusion processing; Based on the regional status categories and regional spatial positioning information in the evidence status map, image spatial annotations are written and status annotations are overlaid on the intermediate set of regional fusion processing to generate evidence fusion annotated images. According to the conflict resolution instructions, a re-evidence prompt is written to the conflict state area in the evidence fusion and annotation image, and the evidence state map and conflict resolution instructions are updated in combination with the uncovered area markings.

[0013] Secondly, the present invention provides a multimodal medical image fusion processing system, including an evidence filing module, which collects multimodal medical image data of cases and pathological sampling records and establishes a correlation relationship to obtain standardized case evidence units and sampling record correlation tables; The topology verification module, based on standardized case evidence units and tissue sampling record association tables, establishes a correspondence chain between imaging lesion areas and pathological field blocks and verifies its credibility, generating cross-modal evidence topology chains and mapping credibility markers. The coverage correction module assesses lesion heterogeneity and pathological coverage based on cross-modal evidence topology chains and mapping credibility markers, and performs representativeness bias correction to generate uncovered area markers and a representativeness correction evidence set. The counterfactual adjudication module verifies the counterfactual evidence against the representative corrected evidence set and the uncovered area markers, and performs state adjudication in conjunction with the mapped credible markers, outputting an evidence state diagram and conflict resolution instructions. The constraint fusion update module performs constraint fusion of the evidence chain based on the evidence status diagram and conflict resolution instructions, generates evidence fusion labeled images, provides re-evidence prompts for conflict areas, and updates the evidence status diagram and conflict resolution instructions in combination with the marking of uncovered areas.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the multimodal medical image fusion processing method as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the multimodal medical image fusion processing method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by organizing the topological connection fields according to the hierarchical connection relationship based on the credibility verification chain set to generate a cross-modal evidence topological chain, and generating a mapping credibility marker, the continuous organization and credibility classification of the correspondence between the image lesion area and the pathological field block are realized; by implementing representative bias correction and reorganizing effective correction evidence to generate a representative correction evidence set, the partition-level coverage bias correction is realized, and the treatment of evidence fusion labeled images is improved in terms of targeting and evidence closure. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a multimodal medical image fusion processing method.

[0019] Figure 2 This is a schematic diagram of a multimodal medical image fusion processing system.

[0020] Figure 3 A flowchart for constructing and verifying the credibility of candidate corresponding chains.

[0021] Figure 4 Heatmap of average coverage error improvement corresponding to coverage offset layering and credibility level determination results.

[0022] Figure 5 A scatter plot showing the targeted effects of representative bias correction measures.

[0023] Figure 6 A comparison chart of targeted hit rates for handling offset layering. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 This is one embodiment of the present invention, which provides a multimodal medical image fusion processing method, including the following steps: S1. Collect multimodal medical imaging data and pathological sampling records of cases and establish correlations to obtain standardized case evidence units and sampling record correlation tables.

[0028] The multimodal medical imaging data of the case includes case identification, multimodal tomographic images, pathological full-view images, acquisition timestamps, image spatial orientation parameters, image spatial resolution parameters, and pathological full-view image coordinate information.

[0029] Pathological sampling records include case identification, sampling number, paraffin block number, slide number, slide orientation marking, sampling site, and layer information.

[0030] The multimodal medical imaging data and pathological sampling records of cases are processed by time information correction, name field unification and spatial benchmark standardization to form a standardized case dataset.

[0031] Furthermore, the acquisition timestamps and recording times of the multimodal medical imaging data and pathological sampling records of the cases are read and time information correction is performed. Different time formats are converted into a unified time format and time alignment is completed according to the case identifier to obtain time-corrected multimodal medical imaging data and pathological sampling records. The name field is unified for the time-corrected multimodal medical imaging data and pathological sampling records, mapping synonymous field names to unified field names and maintaining consistent field values. Based on the spatial normalization method of DICOM spatial metadata, spatial reference standardization processing is performed on the multimodal medical imaging data and pathological sampling records of the cases after name field unification. Based on the coordinate system transformation method of affine transformation matrix, the spatial direction parameters, spatial resolution parameters and pathological full-view image coordinate information are transformed to a unified coordinate reference to form a standardized case dataset.

[0032] Based on the standardized case dataset, a matching index is established between the sampling number and the multimodal medical image data of the case. The multimodal medical image data of the case and the pathological sampling records are associated, encapsulated, and written into the association field to generate a standardized case evidence unit and a sampling record association table.

[0033] Furthermore, in the standardized case dataset, multimodal medical imaging data and pathological sampling records are aggregated by case identifier. Sampling number, paraffin block number, slide number, sampling site, torsional information, and the acquisition timestamp and spatial reference fields corresponding to the multimodal medical imaging data are read. A sampling matching index is established between sampling numbers and multimodal medical imaging data based on consistent case identifiers, matching sampling numbers, consistent sampling sites, and temporal proximity. Based on the sampling matching index, successfully matched multimodal medical imaging data and pathological sampling records are associated and encapsulated according to a unified field structure. The sampling matching relationship and number chain information are written into the association field to generate standardized case evidence units. An index mapping from sampling records to standardized case evidence units is established based on the encapsulation position and sampling matching relationship in the standardized case evidence units, generating a sampling record association table.

[0034] It should be noted that the standardized case evidence unit and the pathology sampling record association table are cross-modal evidence organization and indexing carriers, used to carry the association between multimodal medical imaging data of cases and pathology sampling records, so that the subsequent construction of cross-modal evidence topology chain and pathology coverage assessment have stable input and reduce the risk of mismatch.

[0035] S2. Based on standardized case evidence units and material sampling record association tables, establish a correspondence chain between imaging lesion areas and pathological field blocks and verify its credibility, and generate cross-modal evidence topology chains and mapping credibility markers.

[0036] The location information of the imaging lesion area and the coordinate information of the pathological field of view are extracted from the standardized case evidence unit. The candidate pairing relationship between the imaging lesion area and the pathological field of view is established by combining the sampling record association table, forming a candidate dataset of lesion field of view.

[0037] Furthermore, the location information of the lesion region in the imaging and the coordinate information of the pathological field of view are read from the standardized case evidence unit. The encapsulation position and associated record under the same sampling number chain are located according to the sampling record association table. The location information of the lesion region in the imaging and the coordinate information of the pathological field of view are written into the same candidate pairing calculation unit according to a unified spatial reference. Based on the sampling number, paraffin block number, slide number, sampling site and slice information in the sampling record association table, the number chain consistency constraint and slice proximity constraint are applied to the candidate pairing calculation unit to obtain the candidate pairing relationship that satisfies the sampling association constraint. The candidate pairing relationship that satisfies the sampling association constraint is encapsulated with the corresponding location information of the lesion region in the imaging, the coordinate information of the pathological field of view and the index field of the sampling record association table to form a candidate dataset of lesion fields of view.

[0038] It should be noted that the imaging lesion area refers to the image area used to represent the spatial location and extent of the lesion in the multimodal medical imaging data of the case; the pathological field of view block refers to the local field of view area in the pathological full field of view image that is divided according to the coordinate information of the pathological full field of view image and used to carry pathological tissue information.

[0039] Sampling association constraints refer to the constraints imposed on the consistency and hierarchical restrictions on the candidate pairing relationship between imaging lesion areas and pathological field blocks based on the sampling number, paraffin block number, slide number, sampling site and layer information in the sampling record association table. For example, only candidate pairing relationships that are consistent in sampling site, adjacent in layer and satisfy the same sampling number chain are retained.

[0040] Based on the relationship between the lesion field candidate dataset and the sampling record association table, a hierarchical candidate correspondence relationship between the imaging lesion region and the pathological field block is established, and a candidate correspondence chain set is generated.

[0041] Furthermore, based on the candidate pairing relationship identifiers, imaging lesion region location information, and pathological field block coordinate information in the lesion field candidate dataset, and combined with the sampling number relationship, paraffin block number, slide number, sampling site, and layer information in the sampling record association table, the candidate pairing relationships are grouped according to the same sampling number chain; within each sampling number chain, the pathological field blocks are arranged in hierarchical order from paraffin block number to slide number, and are hierarchically linked with the corresponding imaging lesion region to establish a hierarchical candidate correspondence relationship from the imaging lesion region to the pathological field block; the hierarchical candidate correspondence relationship, together with the sampling number, paraffin block number, slide number, layer information, imaging lesion region location information, and pathological field block coordinate information, is encapsulated in a unified chain field to generate a candidate correspondence chain set.

[0042] Perform number continuity verification, slice direction consistency verification, and spatial adjacency verification on the candidate corresponding chain set to form a set of credibility verification chains.

[0043] Furthermore, within the candidate corresponding chain set, the sampling number, paraffin block number, slide number, slide orientation marker, tomographic information, imaging lesion location information, and pathological field coordinate information are expanded according to the intra-chain order of each candidate corresponding chain. First, the continuity of the sampling number, paraffin block number, and slide number is checked based on the chain-like order relationship to obtain the continuity check conclusion. Then, the continuity check conclusion is compared with the slide orientation marker and the arrangement direction in the hierarchical candidate corresponding relationship to perform the slide orientation consistency check and obtain the slide orientation consistency check conclusion. Finally, the continuity of the sampling number is combined with the sequence order of the slide number. The verification conclusions, slide orientation consistency verification conclusions, imaging lesion area location information, pathological field block coordinate information, and tomographic information undergo spatial adjacency verification to obtain spatial adjacency verification conclusions. Finally, the number continuity verification conclusions, slide orientation consistency verification conclusions, and spatial adjacency verification conclusions are written into the verification fields of the corresponding candidate chain according to the candidate corresponding chain identifier. These are then encapsulated with the aforementioned sample number, paraffin block number, slide number, slide orientation marker, tomographic information, imaging lesion area location information, and pathological field block coordinate information in a unified chain field structure to form a set of credibility verification chains.

[0044] Based on the credibility verification chain set, the topological connection fields are organized according to hierarchical connection relationships to generate cross-modal evidence topological chains. Based on the verification conclusions of each candidate corresponding chain in the credibility verification chain set, a mapping credibility tag is generated.

[0045] Furthermore, within the credibility verification chain set, candidate corresponding chains are grouped according to the sampling number relationship. The wax block number, slice number, slice direction mark, layer information, image lesion area location information, pathological field coordinate information, and verification fields of each candidate corresponding chain are read. Based on the hierarchical connection relationship, the preceding and subsequent connection positions are determined, and the topological connection fields are organized. The topological connection fields, together with the corresponding candidate corresponding chains, are encapsulated in a unified chain field structure order to generate a cross-modal evidence topological chain. At the same time, the number continuity verification conclusion, slice direction consistency verification conclusion, and spatial adjacency verification conclusion are extracted from the credibility verification chain set. The conclusions are merged and graded according to the candidate corresponding chain identifier and the topological connection fields (for example, the number continuity verification conclusion, slice direction consistency verification conclusion, and spatial adjacency verification conclusion are summarized according to the candidate corresponding chain identifier, and the grade is determined in combination with the continuity of the topological connection fields: three items pass the judgment as a high credibility mapping, a single item abnormality is judged as a mapping to be reviewed, and multiple abnormalities or broken topological connection fields are judged as a low credibility mapping). The judgment results are written into the label field to generate a mapping credibility label.

[0046] It should be noted that hierarchical connection refers to the corresponding connection relationship formed by connecting the imaging lesion area and the pathological field of view block step by step according to the sampling level within the same case, based on the sampling number relationship, the slice direction relationship and the spatial adjacency relationship.

[0047] The verification conclusion of the candidate corresponding chain refers to the judgment result obtained after performing number continuity verification, slice direction consistency verification, and spatial adjacency verification on each candidate corresponding chain. It is used to characterize the credibility status of the corresponding candidate corresponding chain and is written into the verification field. For example, the verification conclusion is "number continuity verification passed, slice direction consistency verification passed, spatial adjacency verification abnormal".

[0048] like Figure 4 As shown, after jointly statistically analyzing the coverage offset stratification and credibility level determination results, the average coverage error improvement corresponding to different credibility level determination results shows an overall increasing trend as the coverage offset stratification increases. Among them, the average coverage error improvement corresponding to the mapping to be reviewed and the low credibility mapping is higher under the higher coverage offset stratification, while the average coverage error improvement corresponding to the high credibility mapping is relatively lower and changes more gradually. This indicates that after completing the representative deviation correction and forming a representative correction evidence set, the present invention can prioritize the correction resources to the lesion heterogeneous partition with more prominent coverage deviation and higher credibility risk based on the mapping credibility label and the size of the coverage offset, thereby more effectively realizing partition-level coverage deviation correction and providing a more targeted evidence basis for subsequent evidence state diagram generation and conflict handling instruction output.

[0049] S3. Based on the cross-modal evidence topology chain and the mapping credibility marker, assess the heterogeneity of lesion partitioning and pathological coverage, and implement representativeness bias correction to generate a set of evidence for uncovered areas and representativeness correction.

[0050] Based on the cross-modal evidence topology chain and mapping credibility markers, credibility constraint analysis is performed on the corresponding nodes of the imaging lesion region and the pathological field of view, and the spatial coverage is organized to form a lesion coverage analysis dataset.

[0051] Furthermore, based on the topological connection field, candidate corresponding chain identifier, image lesion region location information, pathological field of view coordinate information, sample number relationship, and torsional information in the cross-modal evidence topological chain, the corresponding nodes of the image lesion region and the pathological field of view are hierarchically expanded and located to obtain the corresponding node location results. Based on the corresponding node location results and the candidate corresponding chain identifier, label field, and confidence level judgment results in the mapping confidence mark, the equi-join and conditional filtering methods are used (e.g., equi-joining the corresponding node location results with the mapping confidence mark according to the candidate corresponding chain identifier, and judging the confidence level as high confidence). (For mappings or mappings to be verified where the marked field is not marked as a broken topological connection field, condition filtering is performed) Confidentiality constraint parsing is performed on the corresponding nodes to obtain the confidence constraint parsing results and the set of corresponding nodes that satisfy the confidence constraints; based on the set of corresponding nodes that satisfy the confidence constraints and the corresponding node positioning results, the spatial coverage boundary, coverage layer order, and coverage overlap relationship of the corresponding image lesion area of ​​the pathological field block are organized to obtain the spatial coverage range organization results; the corresponding node positioning results, confidence constraint parsing results, and spatial coverage range organization results are associated and encapsulated according to a unified field structure to form a lesion coverage analysis dataset.

[0052] The image lesion regions in the lesion coverage analysis dataset are partitioned according to regional difference characteristics, and partition boundaries and partition identifiers are established to generate a lesion heterogeneity partition set.

[0053] Furthermore, based on the image lesion region location information, spatial coverage range organization results, and credibility constraint analysis results in the lesion coverage analysis dataset, regional difference features are obtained within a unified spatial reference using a texture feature extraction method based on the gray-level co-occurrence matrix. Scale normalization is performed on the regional difference features, and K-means clustering combined with region growing is used to partition the image lesion region under spatial adjacency constraints. Connectivity analysis and morphological merging are performed on the partitions to remove scattered small regions and retain continuous regions. The partition boundaries are extracted along the outer contour of the continuous regions, and unique partition identifiers are written according to the spatial order in the image lesion region location information. The partition boundaries and partition identifiers are associated and encapsulated according to a unified field structure to generate a lesion heterogeneity partition set.

[0054] It should be noted that regional difference features refer to the numerical features used to distinguish different regions, including gray-level co-occurrence matrix contrast, gray-level co-occurrence matrix energy, gray-level co-occurrence matrix homogeneity, local entropy, and gradient magnitude.

[0055] Lesion heterogeneity partition sets refer to a set of regions formed by dividing the imaging lesion area according to regional differences and with partition boundaries and partition markers. It is used to conduct pathological coverage assessment at the partition granularity, avoid the overall statistics from masking local uncovered areas and improve the localization accuracy of uncovered area markers.

[0056] Based on the lesion heterogeneity partition set and lesion coverage analysis dataset, the coverage of each partition by the pathological field block is statistically analyzed, and pathological coverage is evaluated by combining the mapping confidence label, generating the label of the uncovered area.

[0057] The expression for pathology coverage assessment is: ; in, It is the first Pathological coverage of heterogeneous regions of individual lesions; It is the lesion heterogeneity partition number, used to distinguish different lesion heterogeneity partitions; It is the first The regional extent of heterogeneous lesion partitions; It is the first A voxel point in a heterogeneous region of a lesion; It is the pathological field block number, used to distinguish different pathological field blocks involved in coverage assessment; Is participating in the The total number of pathological field blocks assessed by the heterogeneous zonal coverage of each lesion; It is the first The coverage area of ​​each pathological field block after being mapped to a unified space; It is a very small positive number, used to prevent the denominator from being zero; It is the first The confidence weight of each pathological field block is obtained by converting the mapping confidence label, and the value ranges from 0 to 1.

[0058] Furthermore, based on the partition identifiers and boundaries of the lesion heterogeneity partition set, the image lesion region location information, pathological field block coordinate information, and spatial coverage range processing results in the lesion coverage analysis dataset are partitioned and aggregated to establish the coverage statistical relationship between pathological field blocks and lesion heterogeneity partitions. Based on the candidate corresponding chain identifiers, label fields, and confidence level judgment results in the mapping confidence markers, the coverage statistical relationship is subjected to validity screening and confidence weight assignment (e.g., the coverage statistical relationship is equi-joined with the mapping confidence markers according to the candidate corresponding chain identifiers, and labels marked as broken topological connections are removed). The statistical relationship of abnormal coverage conclusions is verified, and a higher or lower confidence weight is assigned to the high confidence mapping or the mapping to be reviewed according to the confidence level judgment result, forming a statistical set of partition coverage for participating in pathological coverage assessment; the pathological coverage rate of each lesion heterogeneity partition is calculated according to the expression of pathological coverage assessment, and the pathological coverage rate is written into the coverage assessment field according to the lesion heterogeneity partition number; based on the pathological coverage rate and the coverage judgment threshold, each lesion heterogeneity partition is judged as uncovered or low coverage, and the partition identifier and spatial location corresponding to the judgment are written into the label field to generate uncovered area label.

[0059] It should be noted that the coverage threshold is set based on the historical statistical distribution of pathological coverage of each heterogeneous lesion within the lesion heterogeneity partition set and the consistency results of clinical review, through stratified statistical analysis and cross-validation. An exemplary value range is 0.60 to 0.90.

[0060] Representative bias correction is performed based on the mapping credibility markers, the heterogeneous lesion partition set, and the uncovered area markers, and the effective correction evidence is reorganized to generate a representative correction evidence set.

[0061] Furthermore, based on the credibility level judgment results in the mapping credibility markers and the candidate corresponding chain identifiers, credibility constraints are applied to the partition identifiers of the heterogeneous partition set of lesions corresponding to the uncovered area markers (e.g., after performing equivalent matching between the partition identifier and the uncovered area markers, only partition identifiers with a credibility level judgment result of high credibility mapping or mapping to be reviewed and whose marker fields are not marked as broken topological connection fields are retained), to obtain representative deviation correction target partitions; based on the partition boundaries, image lesion area location information, and pathological field block coordinate information of the representative deviation correction target partitions, the offset between the coverage centroid of the pathological field block and the geometric centroid of the partition is calculated and offset constraints are formed; based on the offset constraints, the credibility level judgment results in the mapping credibility markers, and the spatial coverage range organization results in the lesion coverage analysis dataset, the pathological field block coverage relationship is reweighted and resampled to obtain effective correction candidate relationships; based on the effective correction candidate relationships, the corresponding image lesion area location information, pathological field block coordinate information, partition identifier, credibility level judgment results, and correction weights are written into the evidence field and associated and encapsulated, and the effective correction evidence is reorganized to generate a representative correction evidence set.

[0062] It should be noted that valid correction evidence refers to associated evidence that still meets the mapping credibility label constraint after representativeness bias correction and can stably characterize the lesion heterogeneity partition coverage status, such as "the association record of image lesion area location information and pathological field block coordinate information with partition identifier K3, credibility level judgment result of high credibility mapping and correction weight written".

[0063] like Figure 5 As shown, after completing the representative deviation correction and forming a representative correction evidence set, when the correlation between coverage offset and coverage error improvement is displayed, the overall trend line shows a continuous increase relative to the zero improvement baseline, indicating that the correction magnitude generated by the representative deviation correction is synchronously enhanced when the coverage offset increases. At the same time, the mapping to be reviewed and the low-confidence mapping form a more obvious cluster in the region with higher coverage error improvement, while the distribution of the high-confidence mapping is relatively convergent. This shows that the aforementioned processing can concentrate the correction resources on the partitions with more prominent deviations, thereby improving the pertinence of subsequent regional state category determination and conflict handling instruction generation.

[0064] Coverage offset refers to the offset between the coverage centroid of the pathological field block and the geometric centroid of the representativeness deviation correction target area.

[0065] like Figure 6As shown, when further statistically analyzing the targeted hit rate of treatment after stratifying coverage offsets, the targeted hit rate of the present invention's scheme is higher than that of the static mapping baseline scheme in each coverage offset stratum, and it maintains a high level even in the stratum with higher coverage offsets. This indicates that after completing the representative deviation correction and forming a representative correction evidence set, the present invention's scheme can more accurately target subsequent review, supplementary evidence collection, and re-evidence collection requirements to the heterogeneous lesion partitions with more prominent coverage offsets. This demonstrates that the combined use of mapping credibility markers, uncovered area markers, and representative correction evidence sets not only improves the partition-level coverage offset correction capability but also enhances the targeted treatment in the evidence state diagram and conflict handling instruction generation stages.

[0066] The static mapping baseline scheme is as follows: the correspondence between the imaging lesion area and the pathological field block is established only based on a single pairing or static mapping, and the pathological coverage is evaluated by overall statistics or coarse-grained regional statistics, without performing the comparison scheme of number continuity check, slice orientation consistency check, spatial adjacency check, and representativeness deviation correction.

[0067] The targeted hit rate refers to the proportion of time when the evidence status map and conflict resolution instructions accurately match and hit the target areas with prominent coverage deviations, pending review, conflicts, or no coverage.

[0068] S4. Conduct counterfactual evidence verification on the representative corrected evidence set and the uncovered area markers, and combine the mapped credible markers to make a status decision, outputting an evidence status diagram and conflict resolution instructions.

[0069] Based on the representative correction evidence set and the uncovered area markers, local perturbations that preserve structural boundaries are performed on the imaging lesion areas and pathological field blocks, and the changes in judgment before and after the perturbation are recorded to generate a counterfactual verification record set.

[0070] Furthermore, based on the partition identifiers, image lesion region location information, pathological field block coordinate information, confidence level judgment results, and correction weights in the representative correction evidence set, combined with the partition identifiers and spatial locations in the uncovered area markers, the image lesion region and pathological field block are located and a counterfactual perturbation target set is formed. Based on the counterfactual perturbation target set, a local affine transformation method with boundary mask constraints and a local intensity perturbation method are used to perform local perturbations that preserve structural boundaries and form a perturbation execution record. Based on the perturbation execution record, the judgment states before and after are obtained (e.g., "the judgment state before perturbation is consistent with the evidence, and the judgment state after perturbation is still consistent with the evidence" or "the judgment state before perturbation is consistent with the evidence, and the judgment state after perturbation becomes conflicting with the evidence") and a judgment change record is formed. The counterfactual perturbation target set, perturbation execution record, and judgment change record are associated and encapsulated to generate a counterfactual verification record set.

[0071] It should be noted that the correction weights are based on the confidence level determination results in the mapping confidence label, the pathological coverage assessment results, and the statistical distribution of the coverage offset of the representative deviation correction target partition. The initial weights are determined by the entropy weight method and the parameter calibration is set by combining the K-fold cross-validation method.

[0072] Based on the counterfactual verification record set, the consistency and conflict of evidence in each area of ​​evidence to be verified are determined, and the missing status information is supplemented by the marking of uncovered areas to generate a set of evidence credibility conclusions.

[0073] Furthermore, based on the partition identifiers, image lesion area location information, pathological field block coordinate information, disturbance execution records, and judgment change records in the counterfactual verification record set, the evidence area to be verified is grouped according to the partition identifiers and spatial locations to form a judgment unit for the evidence area to be verified. The judgment stability and judgment reversal are determined based on the judgment states before and after disturbance in the judgment unit for the evidence area to be verified (e.g., if the judgment states before and after disturbance are consistent, it is determined as stable with no judgment reversal; if the judgment states before and after disturbance are inconsistent, it is determined as unstable with judgment reversal), resulting in evidence consistency judgment results and evidence conflict judgment results. Missing state information is supplemented for the judgment unit for the evidence area to be verified based on the partition identifiers and spatial locations in the uncovered area markers. The evidence consistency judgment results, evidence conflict judgment results, and missing state information are associated and encapsulated according to a unified field structure to generate a set of evidence credibility conclusions.

[0074] It should be noted that the area of ​​evidence to be verified refers to the area of ​​the imaging lesion and the corresponding area of ​​the pathological field of view that needs to be determined based on the counterfactual verification record set for evidence consistency and evidence conflict.

[0075] Missing status information refers to status information that is supplemented based on the uncovered area markers and is used to characterize the existence of uncovered or insufficiently covered areas in the corresponding regions.

[0076] Based on the mapping of credible tags and the set of evidence credibility conclusions, the execution status adjudication and disposal rules are matched, and the evidence status diagram and conflict disposal instructions are output.

[0077] Furthermore, based on the candidate corresponding chain identifier, tag field, and credibility level judgment result in the mapped credibility mark, it is associated and merged with the partition identifier, spatial location, evidence consistency judgment result, evidence conflict judgment result, and missing status information in the evidence credibility conclusion set according to the candidate corresponding chain identifier and partition identifier to form a status adjudication judgment unit; the status adjudication judgment unit performs status adjudication on each evidence area to be verified, and writes the evidence consistency, evidence conflict, pending review, and uncovered status into the area status field to form an area status set; the area status set is matched with the handling rules according to the status category, specifically: the review, re-examination, retention, and supplementary evidence collection requirements are written into the instruction field, and the area status field and spatial location in the area status set are encapsulated together with the instruction field according to a unified field structure to output the evidence status diagram and conflict handling instructions.

[0078] S5. Based on the evidence status diagram and conflict resolution instructions, perform evidence chain constraint fusion, generate evidence fusion labeled image, provide re-evidence prompts for conflict areas, and update the evidence status diagram and conflict resolution instructions in combination with the marking of uncovered areas.

[0079] Based on the evidence status diagram and conflict resolution instructions, the status category and resolution requirements of each area are read, and the corresponding evidence chain constraint rules are matched according to the status category to form an evidence chain constraint fusion task set.

[0080] Furthermore, based on the regional status field, spatial location, and candidate corresponding chain identifier in the evidence status diagram, as well as the instruction field and handling requirements in the conflict handling instructions (such as retaining current evidence, performing review and verification, supplementing evidence collection, re-collecting evidence, or marking conflict areas), the corresponding regions at the same spatial location are associated and located to form regional status handling reading units; the status categories and handling requirements of each region are extracted based on the regional status handling reading units, and the corresponding evidence chain constraint rules are matched in the evidence chain constraint rule table according to the status category to form regional rule matching results; the regional status handling reading units and regional rule matching results are associated and encapsulated according to a unified task field structure to form an evidence chain constraint fusion task set.

[0081] It should be noted that the chain of evidence constraint rules refer to the set of rules used to convert regional state categories and disposal requirements into chain of evidence constraint fusion tasks. These rules are set according to the correspondence between state categories in the evidence state diagram and disposal requirements in the conflict disposal instructions.

[0082] Based on the evidence chain constraint fusion task set, the multimodal medical image data of the case are called, and evidence chain constraint fusion processing is performed on each image lesion area to form an intermediate set of regional fusion processing.

[0083] Furthermore, candidate corresponding chain identifiers, spatial locations, status categories, treatment requirements, and evidence chain constraint rules are read from the evidence chain constraint fusion task set. The lesion region to be processed is located according to the candidate corresponding chain identifier and spatial location, and a region processing index is generated. Based on the region processing index, the corresponding case identifier, acquisition timestamp, and image content under a unified spatial reference in the multimodal medical image data of the case are called to form the region fusion processing input. The status category, treatment requirements, and candidate corresponding chain identifier in the region fusion processing input are mapped to fusion parameters and spatial constraints. Evidence reconstruction and constraint updates are performed on the lesion region under a unified spatial reference to form a fusion processing record and a fused image region. The region processing index, fusion processing record, and fused image region are associated and encapsulated to form a region fusion processing intermediate set.

[0084] Based on the regional status categories and regional spatial positioning information in the evidence status map, image spatial annotations are written and status annotations are overlaid on the intermediate set of regional fusion processing to generate evidence fusion annotated images.

[0085] Furthermore, based on the regional status category and regional spatial positioning information in the evidence status map, the spatial range and pixel position of each image lesion region are located in the intermediate set of regional fusion processing. The regional spatial positioning information is converted into the labeled coordinate boundary consistent with the intermediate set of regional fusion processing and written into the image spatial label writing field. According to the regional status category, the written image spatial labels are assigned status label values ​​and superimposed on the corresponding regions of the intermediate set of regional fusion processing according to their spatial positions, keeping the fused image content of non-labeled regions unchanged. After completing the image spatial label writing and status label superimposition, the fused image content and status label layer are encapsulated in a unified image structure to generate the evidence fusion labeled image.

[0086] It should be noted that the regional status categories include regional evidence consistency, regional evidence conflict, regional pending review, and regional uncovered; conflicting regions refer to the imaging lesion regions that are determined to be in regional evidence conflict in the evidence status map and require the execution of conflict resolution instructions.

[0087] Regional spatial positioning information refers to the coordinates, boundaries, and torsional information used to determine the location and extent of an imaging lesion area or a fused imaging area under a unified spatial reference.

[0088] According to the conflict resolution instructions, a re-evidence prompt is written to the conflict state area in the evidence fusion and annotation image, and the evidence state map and conflict resolution instructions are updated in combination with the uncovered area markings.

[0089] Furthermore, the candidate corresponding chain identifier, regional spatial positioning information, instruction fields, and handling requirements are read from the conflict resolution instruction. The conflict state area is located in the evidence fusion and annotation image. The re-evidence prompt is written into the corresponding annotation field of the conflict state area, and the correspondence between the candidate corresponding chain identifier and the regional spatial positioning information is maintained. Combining the partition identifier and spatial location in the uncovered area marker, the regional status field and marker field in the evidence state map are updated accordingly, and the corresponding content of the uncovered area marker is written into the relevant record of the conflict state area. Based on the updated regional status field and marker field in the evidence state map, the instruction fields and handling requirements in the conflict resolution instruction are updated synchronously, and the evidence state map and conflict resolution instruction are updated.

[0090] This embodiment also provides a file encryption system, including: an evidence archiving module, which collects multimodal medical image data of cases and pathological sampling records and establishes an association relationship to obtain standardized case evidence units and sampling record association tables; The topology verification module, based on standardized case evidence units and tissue sampling record association tables, establishes a correspondence chain between imaging lesion areas and pathological field blocks and verifies its credibility, generating cross-modal evidence topology chains and mapping credibility markers. The coverage correction module assesses lesion heterogeneity and pathological coverage based on cross-modal evidence topology chains and mapping credibility markers, and performs representativeness bias correction to generate uncovered area markers and a representativeness correction evidence set. The counterfactual adjudication module verifies the counterfactual evidence against the representative corrected evidence set and the uncovered area markers, and performs state adjudication in conjunction with the mapped credible markers, outputting an evidence state diagram and conflict resolution instructions. The constraint fusion update module performs constraint fusion of the evidence chain based on the evidence status diagram and conflict resolution instructions, generates evidence fusion labeled images, provides re-evidence prompts for conflict areas, and updates the evidence status diagram and conflict resolution instructions in combination with the marking of uncovered areas.

[0091] This embodiment also provides a computer device applicable to the multimodal medical image fusion processing method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the multimodal medical image fusion processing method proposed in the above embodiment.

[0092] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0093] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the multimodal medical image fusion processing method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0094] In summary, this invention generates a cross-modal evidence topology chain by organizing the topology connection fields according to the hierarchical connection relationship based on the credibility verification chain set, and generates a mapping credibility marker, thereby realizing the continuous organization and credibility classification of the correspondence between the image lesion area and the pathological field block; and generates a representative correction evidence set by implementing representative bias correction and reorganizing effective correction evidence, thereby realizing the correction of zonal coverage bias and improving the treatment of evidence fusion-annotated images and the evidence closure.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multimodal medical image fusion processing method, characterized in that, include: Collect multimodal medical imaging data and pathological sampling records of cases and establish correlations to obtain standardized case evidence units and sampling record correlation tables; Based on standardized case evidence units and tissue sampling record association tables, a correspondence chain between imaging lesion areas and pathological field blocks is established and its credibility is verified, generating cross-modal evidence topology chains and mapping credibility markers. Based on the cross-modal evidence topology chain and mapping credibility markers, lesion heterogeneity partitioning and pathological coverage assessment are performed, and representativeness bias correction is implemented to generate uncovered area markers and representativeness correction evidence sets. Counterfactual evidence verification is performed on the representative corrected evidence set and the uncovered area markers, and the status is adjudicated in combination with the mapped credible markers, outputting the evidence status diagram and conflict resolution instructions; Based on the evidence status diagram and conflict resolution instructions, the evidence chain constraint fusion is performed to generate evidence fusion labeled images, prompts for re-examination of conflict areas are given, and the evidence status diagram and conflict resolution instructions are updated in combination with the marking of uncovered areas.

2. The multimodal medical image fusion processing method as described in claim 1, characterized in that, The multimodal medical imaging data of the case includes case identifier, multimodal tomographic images, pathological full-view images, acquisition timestamp, image spatial orientation parameters, image spatial resolution parameters, and pathological full-view image coordinate information; The pathological sampling record includes case identifier, sampling number, paraffin block number, slide number, slide orientation mark, sampling site and layer information.

3. The multimodal medical image fusion processing method as described in claim 2, characterized in that, The steps to obtain the standardized case evidence units and the tissue sampling record association table are as follows: The multimodal medical imaging data and pathological sampling records of cases were processed by time information correction, name field unification and spatial benchmark standardization to form a standardized case dataset; Based on the standardized case dataset, a matching index is established between the sampling number and the multimodal medical image data of the case. The multimodal medical image data of the case and the pathological sampling records are associated, encapsulated, and written into the association field to generate a standardized case evidence unit and a sampling record association table.

4. The multimodal medical image fusion processing method as described in claim 1, characterized in that, The steps for generating the cross-modal evidence topology chain and mapping trusted markers are as follows: The location information of the imaging lesion area and the coordinate information of the pathological field block are extracted from the standardized case evidence unit, and the candidate pairing relationship between the imaging lesion area and the pathological field block is established by combining the sampling record association table to form a candidate dataset of lesion field. Based on the relationship between the lesion field candidate dataset and the tissue sampling record association table, a hierarchical candidate correspondence relationship between the imaging lesion region and the pathological field block is established, and a candidate correspondence chain set is generated. Perform number continuity verification, slice direction consistency verification, and spatial adjacency verification on the candidate corresponding chain set to form a set of credible verification chains; Based on the credibility verification chain set, the topological connection fields are organized according to the hierarchical connection relationship to generate a cross-modal evidence topological chain. Based on the verification conclusions of each candidate corresponding chain in the credibility verification chain set, a mapping credibility tag is generated.

5. The multimodal medical image fusion processing method as described in claim 4, characterized in that, The steps for generating the uncovered area markers and representative correction evidence set are as follows: Based on the cross-modal evidence topology chain and mapping credibility markers, credibility constraint analysis is performed on the corresponding nodes of the imaging lesion region and the pathological field of view block, and the spatial coverage is organized to form a lesion coverage analysis dataset. The image lesion regions in the lesion coverage analysis dataset are partitioned according to regional difference characteristics, and partition boundaries and partition identifiers are established to generate a lesion heterogeneity partition set. Based on the lesion heterogeneity partition set and lesion coverage analysis dataset, the coverage of each partition by the pathological field block is statistically analyzed, and pathological coverage is evaluated by combining the mapping confidence label, generating the label of the uncovered area; Representative bias correction is performed based on the mapping credibility markers, the heterogeneous lesion partition set, and the uncovered area markers, and the effective correction evidence is reorganized to generate a representative correction evidence set.

6. The multimodal medical image fusion processing method as described in claim 1, characterized in that, The steps for outputting the evidence status diagram and conflict resolution instructions are as follows: Based on the representative correction evidence set and the uncovered area markers, local perturbations that preserve structural boundaries are performed on the imaging lesion areas and pathological field blocks, and the changes in judgment before and after the perturbation are recorded to generate a counterfactual verification record set; Based on the counterfactual verification record set, the consistency and conflict of evidence in each area of ​​evidence to be verified are determined, and the missing status information is supplemented by the marking of uncovered areas to generate a set of evidence credibility conclusions. Based on the mapping of credible tags and the set of evidence credibility conclusions, the execution status adjudication and disposal rules are matched, and the evidence status diagram and conflict disposal instructions are output.

7. The multimodal medical image fusion processing method as described in claim 1, characterized in that, The steps are as follows: Based on the evidence status diagram and conflict resolution instructions, evidence chain constraint fusion is performed to generate an evidence fusion labeled image. A re-evidence collection prompt is given for conflict areas, and the evidence status diagram and conflict resolution instructions are updated in conjunction with the marking of uncovered areas. Based on the evidence status diagram and conflict resolution instructions, the status category and resolution requirements of each area are read, and the corresponding evidence chain constraint rules are matched according to the status category to form an evidence chain constraint fusion task set; Based on the evidence chain constraint fusion task set, the case multimodal medical image data is called, and evidence chain constraint fusion processing is performed on each image lesion area to form an intermediate set of regional fusion processing; Based on the regional status categories and regional spatial positioning information in the evidence status map, image spatial annotations are written and status annotations are overlaid on the intermediate set of regional fusion processing to generate evidence fusion annotated images. According to the conflict resolution instructions, a re-evidence prompt is written to the conflict state area in the evidence fusion and annotation image, and the evidence state map and conflict resolution instructions are updated in combination with the uncovered area markings.

8. A multimodal medical image fusion processing system, based on the multimodal medical image fusion processing method according to any one of claims 1 to 7, characterized in that, include: The evidence filing module collects multimodal medical imaging data and pathological sampling records of cases and establishes correlations to obtain standardized case evidence units and sampling record correlation tables. The topology verification module, based on standardized case evidence units and tissue sampling record association tables, establishes a correspondence chain between imaging lesion areas and pathological field blocks and verifies its credibility, generating cross-modal evidence topology chains and mapping credibility markers. The coverage correction module assesses lesion heterogeneity and pathological coverage based on cross-modal evidence topology chains and mapping credibility markers, and performs representativeness bias correction to generate uncovered area markers and a representativeness correction evidence set. The counterfactual adjudication module verifies the counterfactual evidence against the representative corrected evidence set and the uncovered area markers, and performs state adjudication in conjunction with the mapped credible markers, outputting an evidence state diagram and conflict resolution instructions. The constraint fusion update module performs constraint fusion of the evidence chain based on the evidence status diagram and conflict resolution instructions, generates evidence fusion labeled images, provides re-evidence prompts for conflict areas, and updates the evidence status diagram and conflict resolution instructions in combination with the marking of uncovered areas.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the multimodal medical image fusion processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multimodal medical image fusion processing method according to any one of claims 1 to 7.

Images