A care parameter fitting assessment system that identifies a patient state from an image

Abstract

This invention discloses a nursing parameter adaptation and assessment system for image recognition of patient status, belonging to the field of intelligent nursing technology. It includes: a dual-view acquisition module, which frames the nursing area and a reference image, simultaneously acquires dual-view multispectral polarization images, performs distortion correction and scale / brightness consistency, and generates a dual-view calibration image package; a three-dimensional projection module, which performs specular reflection separation and spectral domain alignment on the dual-view calibration image package and projects it onto a unified three-dimensional surface coordinate system to form a three-dimensional surface layer set; and an evidence traversal module, which performs geodesic equal-area segmentation on the three-dimensional surface layer set, extracts multimodal evidence, performs directional traversal along the dominant direction, and converges abnormal pathways to form an abnormal topology map and abnormal boundaries. This invention, through geodesic equal-area segmentation and multimodal evidence arrangement, realizes the expression of the connectivity structure and expansion direction of abnormal regions.

Description

Technical Field

[0001] This invention relates to the field of intelligent nursing technology, and in particular to a nursing parameter adaptation and assessment system based on image recognition of patient status. Background Technology

[0002] In recent years, clinical nursing image evaluation has gradually evolved from single visible light photographs to a structured assessment approach that integrates stereo vision, multispectral imaging, and polarization imaging. A common practice is to acquire images of the nursing area within the nursing distance, use a ruler or reference sticker to achieve scale calibration, and then obtain relatively stable surface texture through spectral alignment or polarization suppression of reflection. At the same time, stereo matching is used to reconstruct the three-dimensional surface, and geometric quantitative indicators such as area and depth are statistically analyzed on the three-dimensional coordinate system. This allows for a stratified description of conditions such as exudation, perfusion, and skin maceration, providing a basis for nursing treatment and follow-up.

[0003] Existing technologies are still susceptible to projection errors caused by specular reflection, cross-view brightness drift, and surface curvature when adapting to nursing parameters. This results in unstable boundaries of abnormal areas and insufficient comparability of the quantitative data collected from different studies. At the same time, two-dimensional or local three-dimensional discrimination often lacks spatial correlation expression along the direction of tissue changes, making it difficult to unify the direction of perfusion defect expansion, the direction of depression deformation, and the spatial distribution of moist coating into executable site-specific nursing decisions. This leads to inconsistencies and over-conservatism in the generation of dressing selection, replacement cycle, decompression range, and reassessment points. Summary of the Invention

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

[0005] Therefore, the present invention provides a nursing parameter adaptation and evaluation system for image recognition of patient status to solve the problem of inconsistent nursing parameter adaptation caused by unstable boundaries and difficulty in quantifying the orientation of abnormal areas in nursing images.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a nursing parameter adaptation and assessment system for patient status based on image recognition, comprising,

[0008] The dual-view acquisition module frames the nursing area and the reference patch together, simultaneously acquires dual-view multispectral polarization images, and completes distortion correction and scale and brightness consistency to generate a dual-view calibration image package.

[0009] The 3D projection module performs specular reflection separation and spectral domain alignment on the dual-view calibration image package and projects it onto a unified 3D surface coordinate system to form a 3D surface layer set.

[0010] The evidence traversal module divides the 3D surface layer set into geodesic area blocks and extracts multimodal evidence. It then performs directional traversal along the dominant direction and converges abnormal paths to form an abnormal topology map and abnormal boundaries.

[0011] The status recording module maps abnormal boundaries to a three-dimensional surface layer set, extracts status quantification indicators, calculates the continuity of perfusion defects, and forms a patient status record.

[0012] The nursing adaptation module performs nursing parameter rule matrix matching on patient status records and abnormal topology maps, determines site-specific nursing measures and review nodes, and generates a nursing parameter execution list.

[0013] As a preferred embodiment of the nursing parameter adaptation and assessment system for image recognition of patient status described in this invention, the generation of the dual-view calibration image package specifically includes:

[0014] The nursing distance and posture of the fixed imaging component are fixed and the field of view is completed to obtain a dual-view framing state including the nursing area and the reference sticker;

[0015] Multispectral imaging and polarization imaging are triggered simultaneously in dual-view framing mode, and timestamps and location identifiers are recorded to obtain dual-view multispectral polarization raw frame groups.

[0016] The dual-view multispectral polarization raw frame group calls the camera intrinsic parameters and performs distortion correction and epipolar correction on the raw frames of each view to obtain the dual-view geometric correction frame group.

[0017] Scale primitives and grayscale reflection blocks are extracted from the reference patch region in the dual-view geometric correction frame group, and scale unification and brightness normalization are performed on the geometric correction frame group of each view to obtain the dual-view calibration frame group.

[0018] The dual-view calibration frame group is combined with timestamps and part identifiers and encapsulated to obtain a dual-view calibration image package.

[0019] As a preferred embodiment of the nursing parameter adaptation and assessment system for image recognition of patient status described in this invention, wherein: the formation of a three-dimensional surface layer set specifically includes:

[0020] The dual-view calibration image packet is separated into co-polarized frames and cross-polarized frames, and the specular reflection component is separated to obtain a dual-view reflection suppression texture.

[0021] Decode the multispectral channels of the dual-view calibration image package and perform inter-channel registration and spectral domain alignment to obtain a dual-view spectral domain consistent map.

[0022] Dual-view stereo matching is performed on the same band view pairs corresponding to the dual-view spectral domain consistent map to obtain the depth field. Then, a three-dimensional point cloud is generated from the depth field to construct a three-dimensional surface mesh, resulting in a unified three-dimensional surface coordinate system.

[0023] The reflection suppression texture and the dual-view spectral domain consistency map are projected and mapped onto a unified three-dimensional surface coordinate system to form a three-dimensional surface layer set of infusion difference indicator map and wetting coating indicator map.

[0024] As a preferred embodiment of the nursing parameter adaptation and assessment system for image recognition of patient status described in this invention, the extraction of multimodal evidence specifically includes:

[0025] The three-dimensional surface mesh of the three-dimensional surface layer set is parameterized with geodesics and an equal-area block mesh is established to obtain the geodesic equal-area block set;

[0026] Extract depressions and curvature undulations within each block to obtain a block geometry entry set;

[0027] Within each block, spatial abrupt ridge lines of the perfusion difference indicator map are extracted, and the residual high reflectance distribution of the reflectivity suppression texture is also extracted to form a block spectral entry set and a block polarization entry set;

[0028] The block geometric entry set, block spectral entry set, and block polarization entry set are merged into block multimodal evidence entries and arranged into an evidence distribution map on a three-dimensional surface;

[0029] The perfusion mutation trend and curvature fluctuation trend are extracted from the multimodal evidence map and merged to obtain the dominant direction field and the initial block set.

[0030] As a preferred embodiment of the nursing parameter adaptation and evaluation system for image recognition of patient status described in this invention, the formation of the abnormal topology map and abnormal boundaries specifically includes:

[0031] Perform forward and backward directional traversals on the initial block set according to the dominant direction field to obtain the block sequence;

[0032] Connectivity filtering is performed on the block sequence based on the continuity of multimodal evidence items, while maintaining surface connectivity, to generate anomaly pathway sets;

[0033] By labeling the intersection nodes and organizing the branch relationships of the abnormal path set, the abnormal topological skeleton is obtained;

[0034] The abnormal topological skeleton is projected onto a 3D surface and the path envelope is closed to obtain the abnormal boundary.

[0035] By combining the abnormal topology skeleton with the abnormal boundary, an abnormal topology graph is obtained.

[0036] As a preferred embodiment of the nursing parameter adaptation and evaluation system for image recognition of patient status described in this invention, the extraction of status quantification indicators specifically includes:

[0037] The abnormal boundary is mapped to a 3D surface layer set and the abnormal region and surrounding skin zone are defined to obtain a boundary region annotation map.

[0038] Within the boundary region annotation map, measure the area of ​​abnormal regions, measure the maximum depression depth, and extract edge irregularity to obtain geometric quantification results;

[0039] The extent of erythema and maceration was extracted within the peripheral skin zone and compared with the wet coating indicator map to obtain the peripheral quantitative results.

[0040] By summing the geometric quantization results with the surrounding quantization results, a set of state quantization indicators is obtained.

[0041] As a preferred embodiment of the nursing parameter adaptation and assessment system for image recognition of patient status described in this invention, the formation of patient status records specifically includes:

[0042] Within the abnormal region, the perfusion defect zone is extracted along the perfusion difference indicator map and a connected piece of the defect zone is formed to obtain the defect zone structure map;

[0043] The connectivity and fracture distribution of the defect zone are calculated on the defect zone structure map, and the extension direction is extracted to obtain the perfusion continuity results.

[0044] The patient status record is obtained by associating and arranging the set of status quantification indicators, perfusion continuity results and abnormal topology graphs into structured record items.

[0045] As a preferred embodiment of the nursing parameter adaptation and evaluation system for image recognition of patient status described in this invention, the step of performing nursing parameter rule matrix matching between the patient status record and the abnormal topology map specifically involves:

[0046] The nursing parameter rule matrix is ​​preset as a combination of dressing attribute items and nursing action items to obtain the rule matrix table;

[0047] Patient status records are mapped to status slots in a rule matrix table to obtain matching path results;

[0048] The status slots corresponding to the matching path results are associated with the optional dressing attribute entries and nursing action entries in the rule matrix table one by one, and consistency filtering is performed to obtain the nursing parameter matching results.

[0049] As a preferred embodiment of the nursing parameter adaptation and assessment system for image recognition of patient status described in this invention, the determination of site-specific nursing measures and reassessment nodes specifically includes:

[0050] The abnormal boundaries and abnormal pathways are projected onto the three-dimensional surface coordinate system, and the abnormal sub-regions are divided according to the spatial connectivity. At the same time, the main direction of the abnormal pathways is extracted and labeled to the corresponding abnormal sub-regions to obtain a localized partition map with direction labels.

[0051] On the site-specific zoning map, determine the dressing type and change cycle, and at the same time determine the cleaning and rinsing intensity, skin barrier protection range, and decompression and position adjustment range to obtain a set of site-specific nursing measures;

[0052] The distribution of patient status records is linked with the set of site-specific nursing measures to form a reassessment node table;

[0053] By merging the set of location-based nursing measures with the review node table, a list of nursing parameter executions is obtained.

[0054] As a preferred embodiment of the nursing parameter adaptation and evaluation system for image recognition of patient status described in this invention, the nursing parameter rule matrix is ​​formed by arranging nursing guide items, material performance items, and nursing action items in an itemized manner.

[0055] The beneficial effects of this invention are as follows: Using a dual-view calibration image package as a unified entry point, the scale and brightness are consistent through a reference patch, and polarization mirror reflection separation is combined to reduce illumination drift and reflection artifacts across viewing angles and acquisitions, ensuring comparability of multispectral information within a unified three-dimensional surface coordinate system after spectral alignment. Simultaneously organizing perfusion difference indicator maps and wetting coating indicator maps within the three-dimensional surface layer set, the state quantification no longer relies on a single texture or viewpoint. Through geodesic equal-area segmentation and multimodal evidence arrangement, abnormal pathways are traversed and converged along the dominant direction to form an abnormal topology map and abnormal boundaries, realizing the connectivity structure and expansion direction expression of abnormal regions. During the state recording stage, geometric quantification results, peripheral quantification results, and perfusion defect continuity results are structurally bound to the abnormal topology map. In the nursing adaptation stage, a nursing parameter rule matrix is ​​used to complete item consistency screening and placement in conjunction with a location-based zoning map, thereby improving boundary stability, quantification reproducibility, and the location-based operability of the nursing execution checklist, reducing unnecessary frequent changes and re-evaluation arrangements. Attached Figure Description

[0056] 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.

[0057] Figure 1 A flowchart for adapting nursing parameters to an assessment system based on image recognition of patient status.

[0058] Figure 2 Flowchart for generating a 3D surface layer set.

[0059] Figure 3Generate a flowchart for the anomaly topology graph.

[0060] Figure 4 Generate a flowchart for the nursing parameter execution list. Detailed Implementation

[0061] 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.

[0062] 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.

[0063] 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.

[0064] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a nursing parameter adaptation assessment system for patient status based on image recognition, comprising the following steps:

[0065] The dual-view acquisition module frames the nursing area and the reference patch together, simultaneously acquiring dual-view multispectral polarization images and completing distortion correction and scale and brightness consistency to generate a dual-view calibration image package.

[0066] Fix the imaging component within the nursing distance range and lock the shooting posture so that the left and right view imaging fields simultaneously cover the nursing area and the reference sticker. Adjust the focus to make the edge of the reference sticker clearly distinguishable, so that the left and right view framing images simultaneously meet the condition of the nursing area and the reference sticker being in the same frame, thus obtaining a dual-view framing state that includes the nursing area and the reference sticker.

[0067] It should be noted that the nursing distance range is the distance interval between the minimum and maximum distances at which the reference patch is fully displayed simultaneously in both the left and right view imaging fields and the edge of the reference patch is distinguishable under clear imaging conditions.

[0068] In dual-view framing mode, left-view multispectral imaging, right-view multispectral imaging, left-view polarization imaging and right-view polarization imaging are triggered simultaneously. At the moment of triggering, the internal clock of the imaging component generates a timestamp and writes it into the image metadata. At the same time, the nursing terminal selects the nursing site from the site list, generates a site identifier and writes it into the image metadata, thus obtaining a dual-view multispectral polarization raw frame group carrying the timestamp and site identifier.

[0069] The camera intrinsic parameters are used to perform radial distortion correction and tangential distortion correction on the left and right view original frames of the dual-view multispectral polarization original frame group, respectively. Epipolar correction is also performed on the left and right view original frames to ensure that the dual-view imaging meets the epipolar alignment constraint, thus obtaining the dual-view geometrically corrected frame group.

[0070] In the dual-view geometric correction frame group, the reference patch region is located, the scale primitives on the reference patch are identified and the pixel length of the scale primitives is measured. At the same time, the grayscale reflective blocks on the reference patch are identified and the reference grayscale values ​​of the grayscale reflective blocks are calculated. The measurement results of the scale primitives and the reference grayscale values ​​of the grayscale reflective blocks are obtained.

[0071] To further explain, the expression for calculating the reference grayscale value of the grayscale reflective block is:

[0072] ;

[0073] in, This indicates the reference grayscale value of the grayscale reflective block. This indicates the number of pixels within the grayscale reflection block area. Indicates the first grayscale reflection block region The grayscale value of each pixel.

[0074] The scale primitive measurement results are used for scale unification of the dual-view geometric correction frame group, and the grayscale reflective block reference grayscale value is used for brightness normalization of the dual-view geometric correction frame group. This completes the image unification process of the left and right views under the same scale and the same brightness reference, resulting in the dual-view calibration frame group.

[0075] To further explain, brightness normalization normalizes the pixel grayscale values ​​of the dual-view geometric correction frame group according to a reference grayscale value. The expression is:

[0076] ;

[0077] in, This represents the pixel grayscale value after brightness normalization. Indicates the pixel position corresponding to the dual-view geometric correction frame group The grayscale value at that location.

[0078] The dual-view calibration frame group is combined with the timestamp and part identifier and encapsulated. The encapsulation content includes at least the left view calibration frame group, the right view calibration frame group, the timestamp, the part identifier, and the camera intrinsic parameter index information corresponding to the imaging component, to obtain the dual-view calibration image package.

[0079] The 3D projection module performs specular reflection separation and spectral domain alignment on the dual-view calibration image package and projects it onto a unified 3D surface coordinate system to form a 3D surface layer set.

[0080] Within the dual-view calibration image package, the left-view calibration frame group and the right-view calibration frame group are read. Based on the polarization direction markings of polarization imaging, co-polarized frames and cross-polarized frames are separated. Co-polarized frames and cross-polarized frames form polarization pairs for the same scene under the same exposure conditions. These polarization pairs are used to separate the specular reflection component, which is represented as the brightness difference component between the co-polarized and cross-polarized frames. The expression is:

[0081] ;

[0082] in, Indicates pixel position The specular reflection component at that location, Indicates pixel position Gray values ​​of frames with the same polarization Indicates pixel position grayscale values ​​of cross-polarized frames This indicates taking the non-negative result;

[0083] The specular reflection component is subtracted from the same polarization frame to obtain the dual-view reflection suppression texture, and the dual-view reflection suppression texture maintains a one-to-one correspondence with the dual-view calibration image package.

[0084] The channel markers of multispectral imaging are read from the dual-view calibration image package and the multispectral channels are decoded to obtain the left-view multispectral channel image group and the right-view multispectral channel image group. Due to the difference between the beam splitter and the imaging channel, each channel in the multispectral channel image group has a spatial offset. The left-view multispectral channel image group is aligned to the left-view reference channel by inter-channel registration, and the right-view multispectral channel image group is aligned to the right-view reference channel by inter-channel registration. The inter-channel registration is completed by feature point matching and geometric transformation. After the inter-channel registration is completed, the spectral domain alignment is completed by using the channel markers. Spectral domain alignment means that the channel order of the left and right views under the same band marker is consistent and the comparability of the imaging response in the same band is maintained, resulting in a dual-view spectral domain consistency map.

[0085] The same band viewpoint is selected from the dual-view spectral domain consistency map, and dual-view stereo matching is performed using the epipolar alignment constraint after epipolar correction. The dual-view stereo matching calculates the parallax along the epipolar direction and forms a depth field. The depth field converts the positional difference of the same spatial point in the left and right views into distance information. Using the focal length parameters and principal point parameters corresponding to the camera intrinsic index information carried by the dual-view calibration image package, the depth field is back-projected into a three-dimensional point cloud and triangulated to construct a three-dimensional surface mesh. The three-dimensional surface mesh carries the three-dimensional vertex coordinates and the connection relationship of the triangular facets to form a unified three-dimensional surface coordinate system.

[0086] A projection relationship is established for each visible vertex of the 3D surface mesh within a unified 3D surface coordinate system. This projection relationship maps the 3D vertex to the pixel positions of the left-view spectral domain consistency map and the right-view spectral domain consistency map. The projection relationship also maps to the pixel positions of the dual-view reflection suppression texture. The infusion difference indicator map is calculated using the projected multispectral pixel values. The infusion difference indicator map is formed by the ratio of the grayscale of the first band to the grayscale of the second band from the same spectral band viewpoint. The first and second bands are determined by multispectral channel markers, and the expression is:

[0087] ;

[0088] in, Indicates pixel position The perfusion difference indicator value at the location, Indicates pixel position The first band gray value, Indicates pixel position The second band gray value is used to characterize the change in reflectance intensity corresponding to the difference in hemoglobin absorption.

[0089] Wetting and coating indicator maps are calculated by using the dual-view reflection suppression texture after projection and the grayscale relationship between the near-infrared and visible light bands in the multi-spectral channels. The wetting and coating indicator maps record the spatial distribution of wetting and coating in the form of vertices on a unified three-dimensional surface coordinate system. The wetting and coating indicator maps are calculated by the ratio of the grayscale values ​​of the near-infrared and visible light bands. The infusion difference indicator map and the wetting and coating indicator map are organized according to the three-dimensional surface mesh on a unified three-dimensional surface coordinate system to obtain a three-dimensional surface layer set.

[0090] The evidence traversal module divides the 3D surface layer set into geodesic areas and extracts multimodal evidence. It then performs directional traversal along the dominant direction and converges abnormal pathways to form an abnormal topology map and abnormal boundaries.

[0091] The three-dimensional surface mesh, saturation difference indicator map, and wet coating indicator map are read from the three-dimensional surface layer set, and the projection relationship under the unified three-dimensional surface coordinate system is preserved. The projection relationship provides the saturation difference indicator value, wet coating indicator value, and vertex sampling value of the dual-view reflection suppression texture corresponding to the three-dimensional surface mesh vertex, thus obtaining the vertex sampling table.

[0092] When performing geodesic parameterization on a 3D surface mesh, several boundary vertices on the 3D surface mesh boundary are selected as geodesic starting points. The shortest path length from each vertex to the set of boundary vertices is calculated along the edges of the 3D surface mesh. The shortest path length is represented by the geodesic distance, which characterizes the surface distance on the 3D surface. Geodesic parameterization maps the geodesic distance to two-dimensional parameter coordinates and forms a meshed parameter domain. The two-dimensional parameter domain is divided into several parameter units according to the principle of equal area partitioning. Each parameter unit points back to the set of vertices and the set of triangular faces on the 3D surface mesh, resulting in the geodesic equal area block set and the block adjacency table.

[0093] When extracting depressions and curvature undulations block by block within a geodesic area block set, a block adjacency table is used to locate adjacent blocks of each block and define the local neighborhood of the block. The depression is calculated by the normal projection distance between the vertex normal and the mean point of the local neighborhood, and the expression is:

[0094] ;

[0095] in, Represents the vertex in a unified three-dimensional surface coordinate system The amount of depression at that location, Represents vertices The unit normal vector at that location, A three-dimensional coordinate vector representing a vertex. Represents vertices The mean of the three-dimensional coordinate vectors of adjacent vertices within the local neighborhood of the partition. This indicates taking the absolute value.

[0096] The range and root mean square of the concavity of all vertices within a block are used to characterize the variation of concavity and curvature fluctuations, resulting in a block geometry entry set.

[0097] When extracting the spatial abrupt ridge of the perfusion difference indicator map within each block, the vertex sampling table provides the perfusion difference indicator value at the block vertex. The perfusion difference indicator value is piecewise linearly interpolated on the block triangular facet. After interpolation, the gradient magnitude is calculated along the tangent plane of the 3D surface mesh, and the local maximum point set of the gradient magnitude is located. The local maximum point set is connected within the block to form the spatial abrupt ridge, resulting in the block spectral entry set. When extracting the residual high reflectivity distribution of the dual-view reflectivity suppression texture within each block, the vertex sampling table provides the dual-view reflectivity suppression texture sampling value at the block vertex. The block vertex sampling value is normalized, and the number of connected pieces and the area ratio of the connected pieces of the bright vertices within the block are counted to obtain the block polarization entry set.

[0098] When merging the block geometry entry set, block spectral entry set, and block polarization entry set, the amount of concavity undulation, spatial abrupt ridge density, and proportion of residual high reflectivity connected patches in the same block are merged into block multimodal evidence entries. The block multimodal evidence entries are then pointed back to the corresponding block positions of the three-dimensional surface mesh to form an evidence distribution map. After the evidence distribution map completes full coverage of the three-dimensional surface mesh, a multimodal evidence map is obtained.

[0099] When extracting the perfusion mutation direction and curvature undulation direction from the multimodal evidence map, the perfusion mutation direction is obtained by least-squares straight-line fitting of the spatial mutation ridge point set in the tangent plane of the 3D surface mesh. The curvature undulation direction is determined by the direction of the maximum concavity change between adjacent blocks of the block geometric entry set. The perfusion mutation direction and curvature undulation direction are merged into a single direction vector within the same block and written into the block direction record. The block direction record forms the dominant direction field on the 3D surface mesh. After the dominant direction field is formed on the 3D surface mesh, the geometric entry values, spectral entry values, and polarization entry numbers in the block multimodal evidence entries are... The values ​​are normalized to the same numerical range, and the sum of the three normalized results within the same block is used to obtain the block evidence strength value. The block evidence strength value is used to sort the blocks in descending order. Blocks with larger block evidence strength values ​​are determined to be blocks ranked higher. Blocks with the same block evidence strength value are ranked according to the sum of the block evidence strength values ​​of adjacent blocks in the block adjacency table. The block with the larger adjacency sum value is determined to be a block ranked higher. The first few blocks after descending order are formed into the initial block set. The initial block set and the dominant direction field are used together for forward and backward directional traversal and to generate the block sequence.

[0100] When performing forward and backward traversals of the initial block set according to the dominant direction field, the block adjacency table limits the candidate adjacent blocks in the traversal direction. Candidate adjacent blocks are selected through directional consistency filtering of the dominant direction field and through block multimodal evidence item continuity filtering. Block multimodal evidence item continuity is calculated from the item differences of adjacent blocks, and the expression is:

[0101] ;

[0102] in, Indicates block With block The continuity value of evidence between them and Indicates block and The geometric values ​​of the entries, and Indicates block and Spectral entry values, and Indicates block and The polarization entry values, and This indicates a block index.

[0103] Adjacent blocks whose continuity values ​​satisfy the connection condition form a block sequence in traversal order, and the block sequences of all starting blocks are collected to obtain the abnormal path set.

[0104] When annotating the intersection nodes and organizing the branching relationships of the abnormal path set, the block adjacency table is used to identify the intersection relationships of multiple abnormal paths in the same block. The intersection relationships form an intersection node list, and the branching relationships form a path adjacency table. The intersection node list and the path adjacency table together obtain the abnormal topological skeleton. When the abnormal topological skeleton is projected onto the 3D surface, the set of triangular facets on the outer edge of the abnormal path set is enveloped and closed to form a closed boundary line. The closed boundary line is the abnormal boundary. The abnormal topological skeleton and the abnormal boundary are combined to form an abnormal topological graph.

[0105] The status recording module maps abnormal boundaries to a three-dimensional surface layer set, extracts status quantification indicators, calculates the continuity of perfusion defects, and forms a patient status record.

[0106] Vertex mapping is performed between the anomaly boundary and the 3D surface mesh of the 3D surface layer set. The mapping process uses the vertex index of the unified 3D surface coordinate system to label the triangular facets covered by the anomaly boundary as anomaly regions. A ring of blocks is extended outward from the anomaly boundary along the block adjacency relationship table of the 3D surface mesh as the peripheral skin zone. The peripheral skin zone represents the strip-shaped skin region adjacent to the anomaly boundary outside the anomaly boundary. The anomaly region and the peripheral skin zone together form the boundary region annotation map.

[0107] The area of ​​the abnormal region is obtained by summing the areas of the triangular facets in the boundary region annotation map. The reference plane is obtained by least-squares plane fitting of the set of vertices of the surrounding skin zone in the boundary region annotation map. The maximum normal distance between the reference plane and the vertices of the abnormal region is taken as the maximum value to obtain the maximum indentation depth. The edge irregularity is calculated by taking the perimeter of the closed boundary line of the abnormal boundary in a unified 3D surface coordinate system and combining it with the area of ​​the abnormal region to obtain the irregularity value. The calculation relationship between the edge irregularity, the area of ​​the abnormal region, and the perimeter is calculated using the perimeter and area, expressed as:

[0108] ;

[0109] in, This indicates the numerical value of edge irregularity. This represents the perimeter of the closed boundary line. Represents pi (π). Indicates the area of ​​the abnormal region.

[0110] The median of the perfusion difference indicator value was extracted from the perfusion difference indicator map of the peripheral skin zone as the peripheral control value. Connected regions in the peripheral skin zone with perfusion difference indicator values ​​higher than the peripheral control value were marked as erythema ranges, representing areas within the peripheral skin zone where the perfusion difference indicator value was higher than the peripheral control value. The median of the wetting coating indicator value was extracted from the wetting coating indicator map of the peripheral skin zone as the wetting control value. Connected regions in the peripheral skin zone with wetting coating indicator values ​​higher than the wetting control value were marked as immersion ranges, representing areas within the peripheral skin zone where the wetting coating indicator value was higher than the wetting control value. The coverage statistics of the wetting coating indicator map were calculated by dividing the number of vertices in the peripheral skin zone or abnormal area that met the criteria of a wetting coating indicator value higher than the wetting control value into the total number of vertices. The expression is:

[0111] ;

[0112] in, Indicates the coverage ratio. This indicates the number of vertices whose wet coating indicator value is higher than the wet control value. This represents the total number of vertices included in the statistics.

[0113] The geometric quantization results are combined with the surrounding quantization results to obtain the state quantization index set.

[0114] Based on the abnormal region, the set of vertices with perfusion difference indicator values ​​lower than the surrounding control values ​​are extracted from the perfusion difference indicator map and connected according to the three-dimensional surface mesh connectivity to form a band structure. The band structure is labeled as the perfusion defect band, and each connected region of the perfusion defect band is labeled as a connected piece of the defect band, forming a defect band structure map. The number of connected pieces of the defect band is counted on the defect band structure map, and the fracture gaps between the connected pieces are located to form a fracture distribution. The vertex set of the defect band in the defect band structure map is fitted with a least-squares straight line in the tangent plane of the unified three-dimensional surface coordinate system to obtain the extension direction. The extension direction represents the main extension direction of the perfusion defect band in the abnormal region. The defect band structure map, fracture distribution, and extension direction together form the perfusion continuity result.

[0115] The patient status record is obtained by associating and arranging the status quantification index set, perfusion continuity results and abnormal topology map. The association and arrangement binds the status quantification index set to the abnormal boundary label, binds the perfusion continuity results to the abnormal pathway direction, and writes the timestamp and location identifier of the dual-view calibration image package into the structured record item to form the patient status record.

[0116] The nursing adaptation module performs nursing parameter rule matrix matching on patient status records and abnormal topology maps, determines site-specific nursing measures and review nodes, and generates a nursing parameter execution list.

[0117] The nursing guide items, material performance items, and nursing action items are arranged in an itemized manner. The nursing guide items include status classification and corresponding treatment points. The material performance items include dressing absorption capacity, breathability, antibacterial properties, ability to fit curved surfaces, and replacement cycle. The nursing action items include cleaning and rinsing intensity, skin barrier protection range, decompression and positioning adjustment range, and expression method of review nodes. The itemized arrangement arranges the status classification field into status slots and the dressing attribute items and nursing action items into optional item columns, resulting in a rule matrix table and forming a nursing parameter rule matrix.

[0118] The state slots in the rule matrix table receive the set of state quantification indicators and perfusion continuity results from the patient's state records. The state quantification indicator set is mapped to the abnormal area level, maximum indentation depth level, edge irregularity level, erythema extent level, and infiltration extent level, respectively. The perfusion continuity results are mapped to the perfusion defect continuity level while preserving the expansion direction. This mapping process yields the state slot label vector. The state slot label vector is then used to locate the set of rows satisfying the same slot label in the rule matrix table, and the intersection of these rows is calculated to form the matching path result. The expression is:

[0119] ;

[0120] in, This represents the set of candidate rule rows corresponding to the matching path results. Indicates the number of status slots. The rule matrix table represents the condition that satisfies the first... A set of rule rows for labeling each status slot, with symbols This represents the intersection operation of sets.

[0121] The matching path results output a set of candidate rule rows and a status slot label vector.

[0122] The matching path results association rule matrix table contains optional dressing attribute entries and nursing action entries. The candidate rule row set, corresponding to the dressing's absorbency, breathability, antibacterial properties, surface conformation ability, and replacement cycle, forms the dressing candidate set. The candidate rule row set, corresponding to the cleaning and rinsing intensity, skin barrier protection range, decompression, and body position adjustment range, forms the action candidate set. Consistency screening checks the constraints between entries in the dressing candidate set and the action candidate set. The constraints include at least the consistency of dressing absorbency with the coverage ratio of the wet application indicator map, the consistency of dressing surface conformation ability with the maximum indentation depth, the consistency of breathability with the ventilation range of the immersion area, and the consistency of antibacterial properties with the risk of erythema range. Combinations of dressing attribute entries and nursing action entries that satisfy the constraints are retained to obtain the nursing parameter matching results.

[0123] The abnormal boundaries and abnormal paths of the abnormal topology map are projected and labeled in a unified three-dimensional surface coordinate system. The projection label divides the three-dimensional surface mesh into abnormal regions according to the closed areas of the abnormal boundaries and further divides them into abnormal sub-regions according to the spatial connectivity. The main direction of the abnormal path is calculated in each abnormal sub-region and written into the abnormal sub-region direction label to form a localized partition map with direction label.

[0124] To further explain, the main orientation is obtained by taking the spatial point set of the abnormal path within the abnormal sub-region and placing it in the tangent plane of the unified three-dimensional surface coordinate system. The specific calculation method is as follows: First, extract the three-dimensional coordinate set of the block center points covered by the abnormal path in the abnormal sub-region. The three-dimensional coordinates of the block center points are obtained by averaging the three-dimensional coordinates of the vertices of the corresponding triangular facets of the block. Then, perform least-squares plane fitting on the vertex set of the three-dimensional surface mesh of the abnormal sub-region to obtain the tangent plane of the abnormal sub-region, and establish two orthogonal tangent basis vectors on the tangent plane of the abnormal sub-region. Subsequently, project the three-dimensional coordinate set of the block center points onto the tangent plane of the abnormal sub-region to obtain a two-dimensional coordinate point set. Calculate the covariance matrix of the two-dimensional coordinate point set and take the eigenvector corresponding to the largest eigenvalue as the two-dimensional direction of the main orientation. Finally, map the two-dimensional direction back to the tangent plane of the abnormal sub-region to obtain the three-dimensional direction of the main orientation and normalize it.

[0125] The nursing parameter matching results are mapped zone by zone on a site-specific zoning map with directional markings. Dressing type and change cycle correspond to the abnormal area area level and the coverage ratio level of the moist coating indicator map. Cleaning and rinsing intensity corresponds to the immersion range level. The skin barrier protection range is determined by covering the surrounding skin zone along the outer edge of the abnormal boundary. The decompression and positioning adjustment range is determined by covering the expansion direction of the perfusion defect zone and the main direction of the abnormal pathway, thus obtaining a site-specific nursing intervention set. The timestamps and site identifiers from the patient status record are added to the site-specific nursing intervention set, and a review node table is determined according to the review interval table given in the nursing guidelines. The time calculation expression for the review node table is:

[0126] ;

[0127] in, Indicates the time of the re-evaluation. This indicates the timestamp corresponding to the patient's status record. This indicates the re-evaluation interval given in the re-evaluation interval table.

[0128] The site-specific nursing intervention set and the reassessment checklist are merged and arranged. The arrangement content should include at least the site identification, abnormal sub-area marking, dressing type, dressing cycle, cleaning and rinsing intensity, skin barrier protection range, decompression and position adjustment range, and reassessment checkpoint time. The merged arrangement outputs a nursing parameter execution list.

[0129] In summary, this invention uses a dual-view calibration image package as a unified entry point. It achieves scale and brightness consistency through a reference patch and combines polarization mirror reflection separation to reduce illumination drift and reflection artifacts across different viewpoints and acquisitions. This ensures that multispectral information remains comparable within a unified three-dimensional surface coordinate system after spectral alignment. Simultaneously, it organizes perfusion difference indicator maps and wetting coating indicator maps within the three-dimensional surface layer set, making state quantification no longer dependent on a single texture or viewpoint. Through geodesic equal-area segmentation and multimodal evidence arrangement, it directionally traverses along the dominant direction to converge abnormal pathways and form anomaly topology maps and boundaries, realizing the connectivity structure and expansion direction expression of abnormal regions. During the state recording stage, geometric quantification results, peripheral quantification results, and perfusion defect continuity results are structurally bound to the anomaly topology map. In the nursing adaptation stage, a nursing parameter rule matrix is ​​used to complete item consistency screening and, combined with a location-based zoning map, placement is achieved. This improves boundary stability, quantification reproducibility, and the location-based operability of the nursing execution checklist, reducing unnecessary frequent changes and re-evaluation arrangements.

[0130] Example 2, referring to Table 1, is the second embodiment of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the nursing parameter adaptation assessment system for image recognition of patient status are given.

[0131] The experimental subjects were selected from three common nursing sites: Subject A had a sacral and coccygeal pressure ulcer, Subject B had a postoperative incision on the dorsum of the foot, and Subject C had a calf abrasion. Each subject completed two procedures during the same imaging period: one procedure used a full workflow of "dual-view multispectral polarization imaging + dual-view calibration image package + three-dimensional surface layer set + abnormal topology map and abnormal boundary + patient status record + nursing parameter execution checklist", and the other procedure served as a control using a conventional workflow of "dual-view visible light imaging + two-dimensional contour extraction + manual verification". The abnormal area was manually calibrated by two nurses who measured the plane according to the reference scale and recorded the consistent results in a table.

[0132] The implementation process is completed in the following order: The imaging component locks its posture within the nursing distance range, so that the nursing area and the reference patch are in the same frame to form a dual-view framing state. Simultaneously, left-view multispectral imaging, right-view multispectral imaging, left-view polarization imaging, and right-view polarization imaging are triggered and timestamps and location identifiers are written to obtain a dual-view multispectral polarization raw frame group. The camera intrinsic parameters are called to perform distortion removal and epipolar correction on the dual-view multispectral polarization raw frame group to obtain a dual-view geometric correction frame group. The reference patch area is located, the scale primitive pixel length is extracted, and the gray level sum and pixel number of gray level reflective blocks are counted. The gray level reflective block reference gray level value is calculated, and the scale is unified and the brightness is normalized on the dual-view geometric correction frame group. The image is then packaged to obtain a dual-view calibration image package.

[0133] Subsequently, 3D projection and evidence traversal were completed: the dual-view calibration image package was separated into co-polarized and cross-polarized frames, and specular reflection components were obtained according to brightness differences. The specular reflection components were subtracted to obtain dual-view anti-reflection textures. After multi-spectral channel decoding, inter-channel registration and spectral domain alignment were performed to obtain a dual-view spectral domain consistency map. Dual-view stereo matching of viewpoints in the same band was performed to obtain a depth field, generating a 3D point cloud and a 3D surface mesh to form a unified 3D surface coordinate system. The anti-reflection texture and the dual-view spectral domain consistency map were projected onto the unified 3D surface coordinate system to form an infusion difference indicator map and a wetting coating indicator map, which were then organized into a 3D surface layer set. The 3D surface layer set underwent geodesic parameterization and equal-area segmentation. Depressions and curvature undulations, spatially abrupt ridge lines, and residual high reflectivity distributions were extracted block by block and merged into a multimodal evidence map. The blocks were sorted in descending order based on their evidence strength values ​​to obtain the initial block set, and then forward-aligned with the dominant direction field. A reverse-directional traversal converges the abnormal pathway set, completes the annotation of intersection nodes and branch organization to obtain the abnormal topological skeleton, and performs pathway envelope closure to form abnormal boundaries and abnormal topological maps; the abnormal boundaries are mapped to a 3D surface layer set to extract the abnormal area, maximum depression depth, edge irregularity, erythema range, maceration range, and wet coverage ratio, and calculate the continuity of the perfusion defect zone to form a patient status record; nursing guide items, material performance items, and nursing action items are arranged to form a nursing parameter rule matrix, the patient status record is mapped to the status slots of the rule matrix table, and the matching path results are obtained by the intersection of row sets, then the dressing attribute items and nursing action items are associated and consistency screening is performed to obtain the nursing parameter matching results; the abnormal boundaries and abnormal pathways are projected and annotated to obtain a location-based zoning map with direction annotations, the nursing parameter matching results are assigned to each zone and the review node time is determined, and the nursing parameter execution list is output.

[0134] The details are shown in Table 1 below:

[0135] Table 1. Comparison of Dual-View Multispectral Polarization Three-Dimensional Errata Data

[0136] parameter Object A - Sacrococcygeal pressure ulcer - This invention Subject A - Sacrococcygeal pressure ulcer - Control Object B - Dorsal incision of the foot - This invention Subject B - Dorsal incision of the foot - Control Object C - Lower leg abrasion - This invention Subject C - Lower leg abrasion - Control Grayscale reflective pixel count - left view 2500 2500 2500 2500 2500 2500 Grayscale reflection block grayscale sum - left view 325000 322500 310000 305000 345000 340000 Grayscale reflective block reference grayscale value - left view 130 129 124 122 138 136 Grayscale reflective pixel count - right view 2500 2500 2500 2500 2500 2500 Grayscale reflection block grayscale sum - right view 327500 350000 312500 330000 346250 370000 Grayscale reflective block reference grayscale value - right view 131 140 125 132 138.5 148 Mean grayscale of ROI in nursing area - left view 156 170 148.8 162 165.6 184 Brightness normalized ROI mean - left view 1.2 1.3178 1.2 1.3279 1.2 1.3529 Mean grayscale of ROI in nursing area - right perspective 157.2 196 150 175 166.2 210 Luminance normalized ROI mean - right view I_norm 1.2 1.4 1.2 1.3258 1.2 1.4189 Poor brightness consistency between left and right viewing angles 0 6.05 0 0.16 0 4.76 Specular reflection residual rate (%) 1.8 9.5 2.1 8.7 2.4 10.2 Spectral alignment error (px) 0.4 1.6 0.5 1.4 0.6 1.8 Depth field reconstruction RMSE (mm) 0.7 1.9 0.8 2.1 0.9 2.4 Manually calibrate the area of ​​abnormality (mm2). 820 820 460 460 1200 1200 The measured area of ​​the anomaly is A (mm2). 812 870 452 510 1186 1295 Perimeter P (mm) of the abnormal boundary 117 132 89 103 149 173 Edge Irregularity IR 1.3415 1.5937 1.3945 1.6554 1.4896 1.8391 Abnormal area error (%) -0.98 6.1 -1.74 10.87 -1.17 7.92 Total number of vertices in the abnormal region 5400 5400 3800 3800 7200 7200 Number of wet vertices 2400 3100 900 1500 3100 4200 Moisture coverage ratio R 0.4444 0.5741 0.2368 0.3947 0.4306 0.5833 Number of abnormal pathways (number of pathways) 2 7 1 5 3 9 Anomaly Boundaries and Annotations 0.86 0.71 0.88 0.73 0.84 0.69 Time taken per evaluation (s) 42 55 39 51 46 60 Recommended replacement cycle (h) 24 12 24 12 24 12 Review interval (h) 24 12 24 12 24 12

[0137] The data in the table show that, compared with the control, the specular reflection residual rate was significantly reduced in all three groups of objects (e.g., object C decreased from 10.2% to 2.4%), and the wet coverage ratio decreased simultaneously (object C decreased from 0.5833 to 0.4306). This indicates that after the polarization pair was separated from the specular reflection component, the false count of high-reflectivity pseudo-wetting in the wet coating indicator map was reduced. The spectral alignment error decreased to the range of 0.4–0.6 pixels in all three groups of objects, while the depth field reconstruction RMSE decreased to the range of 0.7–0.9 mm. This indicates that the dual-view stereo matching and 3D surface mesh construction are subject to the positive constraints of inter-channel registration and spectral alignment, which reduces the geometric drift of the 3D surface coordinate system.

[0138] The structured localization effect brought about by evidence traversal is reflected in the number of abnormal pathways and the accuracy of abnormal boundaries: the number of abnormal pathways in the three groups of subjects was significantly reduced (from 7 to 2 in subject A, and from 9 to 3 in subject C), and the IoU between abnormal boundaries and expert annotations improved to the range of 0.84–0.88, while the control remained in the range of 0.69–0.73; the abnormal area error was controlled within about 2% in the three groups of subjects, while the control reached the highest at 10.87%; the edge irregularity IR was generally larger and more volatile in the control, reflecting that the two-dimensional contour is more prone to jagged boundaries due to reflection and viewing angle differences. At the nursing parameter execution checklist level, the recommended replacement cycle and re-evaluation interval were shorter in the control (12 hours) and more stable in the process of this invention (24 hours), mainly because the improved wet coverage ratio and abnormal boundary stability resulted in clearer constraints on the selection of nursing action items, thereby reducing overly conservative frequent replacement and re-evaluation arrangements.

[0139] 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 nursing parameter adaptation and assessment system for patient status based on image recognition, characterized in that: include, The dual-view acquisition module frames the nursing area and the reference patch together, simultaneously acquires dual-view multispectral polarization images, and completes distortion correction and scale and brightness consistency to generate a dual-view calibration image package. The 3D projection module performs specular reflection separation and spectral domain alignment on the dual-view calibration image package and projects it onto a unified 3D surface coordinate system to form a 3D surface layer set. The evidence traversal module divides the 3D surface layer set into geodesic area blocks and extracts multimodal evidence. It then performs directional traversal along the dominant direction and converges abnormal paths to form an abnormal topology map and abnormal boundaries. The status recording module maps abnormal boundaries to a three-dimensional surface layer set, extracts status quantification indicators, calculates the continuity of perfusion defects, and forms a patient status record. The nursing adaptation module performs nursing parameter rule matrix matching on patient status records and abnormal topology maps, determines site-specific nursing measures and review nodes, and generates a nursing parameter execution list.

2. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 1, characterized in that: The generation of the dual-view calibration image package specifically involves: The nursing distance and posture of the fixed imaging component are fixed and the field of view is completed to obtain a dual-view framing state including the nursing area and the reference sticker; Multispectral imaging and polarization imaging are triggered simultaneously in dual-view framing mode, and timestamps and location identifiers are recorded to obtain dual-view multispectral polarization raw frame groups. The dual-view multispectral polarization raw frame group calls the camera intrinsic parameters and performs distortion correction and epipolar correction on the raw frames of each view to obtain the dual-view geometric correction frame group. Scale primitives and grayscale reflection blocks are extracted from the reference patch region in the dual-view geometric correction frame group, and scale unification and brightness normalization are performed on the geometric correction frame group of each view to obtain the dual-view calibration frame group. The dual-view calibration frame group is combined with timestamps and part identifiers and encapsulated to obtain a dual-view calibration image package.

3. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 2, characterized in that: The formation of the three-dimensional surface layer set specifically involves: The dual-view calibration image packet is separated into co-polarized frames and cross-polarized frames, and the specular reflection component is separated to obtain a dual-view reflection suppression texture. Decode the multispectral channels of the dual-view calibration image package and perform inter-channel registration and spectral domain alignment to obtain a dual-view spectral domain consistent map. Dual-view stereo matching is performed on the same band view pairs corresponding to the dual-view spectral domain consistent map to obtain the depth field. Then, a three-dimensional point cloud is generated from the depth field to construct a three-dimensional surface mesh, resulting in a unified three-dimensional surface coordinate system. The reflection suppression texture and the dual-view spectral domain consistency map are projected and mapped onto a unified three-dimensional surface coordinate system to form a three-dimensional surface layer set of infusion difference indicator map and wetting coating indicator map.

4. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 3, characterized in that: The extraction of multimodal evidence specifically includes: The three-dimensional surface mesh of the three-dimensional surface layer set is parameterized with geodesics and an equal-area block mesh is established to obtain the geodesic equal-area block set; Extract depressions and curvature undulations within each block to obtain a block geometry entry set; Within each block, spatial abrupt ridge lines of the perfusion difference indicator map are extracted, and the residual high reflectance distribution of the reflectivity suppression texture is also extracted to form a block spectral entry set and a block polarization entry set; The block geometric entry set, block spectral entry set, and block polarization entry set are merged into block multimodal evidence entries and arranged into an evidence distribution map on a three-dimensional surface; The perfusion mutation trend and curvature fluctuation trend are extracted from the multimodal evidence map and merged to obtain the dominant direction field and the initial block set.

5. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 4, characterized in that: The formation of the abnormal topology map and abnormal boundaries specifically includes: Perform forward and backward directional traversals on the initial block set according to the dominant direction field to obtain the block sequence; Connectivity filtering is performed on the block sequence based on the continuity of multimodal evidence items, while maintaining surface connectivity, to generate anomaly pathway sets; By labeling the intersection nodes and organizing the branch relationships of the abnormal path set, the abnormal topological skeleton is obtained; The abnormal topological skeleton is projected onto a 3D surface and the path envelope is closed to obtain the abnormal boundary. By combining the abnormal topology skeleton with the abnormal boundary, an abnormal topology graph is obtained.

6. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 5, characterized in that: The extraction status quantification index is specifically as follows: The abnormal boundary is mapped to a 3D surface layer set and the abnormal region and surrounding skin zone are defined to obtain a boundary region annotation map. Within the boundary region annotation map, measure the area of ​​abnormal regions, measure the maximum depression depth, and extract edge irregularity to obtain geometric quantification results; The extent of erythema and maceration was extracted within the peripheral skin zone and compared with the wet coating indicator map to obtain the peripheral quantitative results. By summing the geometric quantization results with the surrounding quantization results, a set of state quantization indicators is obtained.

7. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 6, characterized in that: The formation of the patient status record specifically includes: Within the abnormal region, the perfusion defect zone is extracted along the perfusion difference indicator map and a connected piece of the defect zone is formed to obtain the defect zone structure map; The connectivity and fracture distribution of the defect zone are calculated on the defect zone structure map, and the extension direction is extracted to obtain the perfusion continuity results. The patient status record is obtained by associating and arranging the set of status quantification indicators, perfusion continuity results and abnormal topology graphs into structured record items.

8. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 7, characterized in that: The specific steps for performing nursing parameter rule matrix matching between patient status records and abnormal topology maps are as follows: The nursing parameter rule matrix is ​​preset as a combination of dressing attribute items and nursing action items to obtain the rule matrix table; Patient status records are mapped to status slots in a rule matrix table to obtain matching path results; The status slots corresponding to the matching path results are associated with the optional dressing attribute entries and nursing action entries in the rule matrix table one by one, and consistency filtering is performed to obtain the nursing parameter matching results.

9. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 8, characterized in that: The specific site-specific nursing measures and reassessment points are as follows: The abnormal boundaries and abnormal pathways are projected onto the three-dimensional surface coordinate system, and the abnormal sub-regions are divided according to the spatial connectivity. At the same time, the main direction of the abnormal pathways is extracted and labeled to the corresponding abnormal sub-regions to obtain a localized partition map with direction labels. On the site-specific zoning map, determine the dressing type and change cycle, and at the same time determine the cleaning and rinsing intensity, skin barrier protection range, and decompression and position adjustment range to obtain a set of site-specific nursing measures; The distribution of patient status records is linked with the set of site-specific nursing measures to form a reassessment node table; By merging the set of location-based nursing measures with the review node table, a list of nursing parameter executions is obtained.

10. The nursing parameter adaptation and assessment system for patient status based on image recognition as described in claim 8, characterized in that: The nursing parameter rule matrix is ​​formed by arranging nursing guide entries, material performance entries, and nursing action entries into a list.

Images