Intelligent wound assessment management system and method for large-area burn

Abstract

The invention relates to the technical field of intelligent diagnosis, in particular to an intelligent wound assessment management system and method for large-area burn, and the system comprises an exudation associated region extraction module, a structural stress concentration recognition module, a tissue boundary disturbance recognition module, an adhesion imbalance path calibration module and a wound assessment management output module. According to the method, the correlation between the edge trend of the exudation area and the texture extension direction is extracted, the longitudinal contact track of the middle-layer tissue boundary in the image sequence is combined, the angle change process of the slippage trend and the local direction information is iteratively identified, and the continuous change of the path offset behavior in the time range is tracked; and extracting spatial distribution of structure response features in a layer region, a correlation tension direction, boundary movement and an exudation trend, completing layer output of region state difference, expanding a space-time correlation range of path change and region difference in an image sequence, and enhancing a linkage capability of a structure path and a tissue state under a continuous image.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent diagnosis, in particular to a large-area burn intelligent wound evaluation management system and method. BACKGROUND

[0002] The technical field of intelligent diagnosis involves the use of artificial intelligence, medical image processing, and information system integration to automatically identify and analyze diseases, pathological states, or physiological indicators. This field mainly includes medical image recognition and analysis, physiological parameter automatic acquisition and modeling, disease prediction model construction, abnormal detection and classification, and covers image acquisition methods, image semantic recognition, diagnostic feature extraction, multi-source data fusion processing, prediction model training and updating mechanisms. Intelligent diagnosis is widely used in clinical auxiliary diagnosis, individualized treatment planning, remote health management, and intelligent early warning, especially in disease screening, postoperative monitoring, and wound evaluation. Traditional burn wound evaluation management refers to the collection, analysis, and recording of morphological parameters and tissue activity information of large-area burn patients, usually relying on manual visual judgment, manual measurement after planar image capture, and indirect sampling analysis of tissue activity indicators.

[0003] Existing technologies use single-frame image capture combined with visual judgment to process image regions, and image expression is limited to planar structures, which cannot extract the region response process in time-continuous images. Region state judgment relies on visual observation and point measurement, lacks support for tissue dynamic change trajectory, cannot provide time sequence feedback in terms of strain propagation, boundary response, and local tension trend, and lacks a basis for comparing region direction trend and continuous shift path in image static samples, making it difficult to express actual spatial shift response caused by internal state changes. Layer expression for multi-source information structure has problems such as boundary ambiguity and information coverage limitations, and there is a risk of lag and judgment bias in region state quantization, tissue hierarchical subdivision, and dynamic tracking. SUMMARY

[0004] To solve the technical problems existing in the prior art, the present application provides a large-area burn intelligent wound evaluation management system and method. The technical solution is as follows:

[0005] On the one hand, a large-area burn intelligent wound evaluation management system is provided, which includes: The exudation correlation region extraction module obtains a differential hierarchical image in a burn wound, extracts a continuous exudation texture region according to the edge curvature characteristics and texture extension direction of the liquid high-reflection region, and obtains a set of superficial exudation correlation patches. The structural stress concentration identification module locates the middle layer tissue boundary trend in the corresponding image area based on the superficial exudation associated patch area set, tracks the longitudinal offset path and the deep layer contact contour in the continuous frames, and obtains a structural traction concentration path sequence; The tissue boundary disturbance identification module compares the contour directionality changes in the two side image frames based on the continuous structural band in the structural traction concentration path sequence, extracts the frame segment with reverse offset trend, and obtains a wound boundary drift disturbance path set; The adhesion imbalance path calibration module observes the offset angle of the cell arrangement main direction and the sliding trend in the continuous frames based on the wound boundary drift disturbance path set, identifies the imbalance offset track, and obtains an adhesion structure imbalance path coordinate set; The wound evaluation management output module analyzes the tissue performance, exudation trend and tension state in the path coverage area based on the adhesion structure imbalance path coordinate set, and obtains a wound multi-source state management layer structure.

[0006] As a further scheme of the present application, the superficial exudation associated patch area set includes a high-reflection edge curvature concentrated area, a texture main direction continuous area, and a direction intersection consistent area; the structural traction concentration path sequence includes a middle layer tissue boundary continuous offset line, a deep layer contact contour path, and a closed trend corresponding area; the wound boundary drift disturbance path set includes a reverse offset continuous frame segment, a bending change contour section, and an edge trend disturbance path; the adhesion structure imbalance path coordinate set includes a sliding offset track coordinate, an arrangement direction deviation area, and a structure offset aggregation path; and the wound multi-source state management layer structure includes a tissue type layer, an exudation expansion trend layer, a boundary connectivity state layer, and a tension propagation distribution layer.

[0007] As a further scheme of the present application, the continuous exudation texture area refers to tracking frame by frame along the texture direction after extracting the liquid reflection edge from the image, screening the areas with consistent and continuous direction in multiple frames, and identifying the time and space continuous trend segments. The deep layer contact contour refers to monitoring the longitudinal movement in the frame sequence after extracting the middle layer boundary, and identifying the path continuously approaching the deep part and forming a closed trend with the upper layer edge.

[0008] As a further scheme of the present application, the contour directionality change refers to tracking the tissue edge trend in the continuous image, analyzing whether there is a reverse offset and trend mutation in the adjacent frames, and whether the change occurs continuously in time. The imbalance offset track refers to extracting the arrangement direction and the sliding direction of the path area, frame by frame analyzing whether they are inconsistent for a long time, and when the deviation phenomenon continuously occurs and is accompanied by the spatial offset of the path in the image, it can be judged as an imbalance offset track.

[0009] As a further scheme of the present application, the exudation associated area extraction module includes: The image hierarchy recognition submodule obtains the differentiated images in the burn wound, extracts the gray scale boundary of the liquid high-reflection area, collects the edge band of the brightness jump, extracts the gray scale direction and distribution trend of the edge pixels, locates the image position between the gray scale breakpoints, and obtains a reflection edge region set; The edge trend analysis submodule extracts the edge curvature direction and texture extension direction based on the reflection edge region set, compares the cross trend in space, compares the trend continuity in the image frame sequence, extracts the frame segment image with consistent cross direction, and obtains a cross trend continuous segment set. The extension area extraction submodule traces the texture area in the image frame sequence based on the cross trend continuous segment set, screens the positions with consistent direction extension, extracts the covered area in the time sequence, locates the corresponding image block, and obtains a superficial exudation associated area set.

[0010] As a further scheme of the present application, the structure stress concentration recognition module comprises: The boundary line extraction processing submodule obtains the corresponding position of the middle layer tissue based on the superficial exudation associated area set, extracts the edge band direction of the gray scale mutation area and the edge pixel sequence along the horizontal axis, traces the edge continuous trend in the frame sequence, locates the corresponding track line position in each frame image, and obtains a middle layer boundary line track set. The longitudinal path change recognition submodule calls the middle layer boundary line track set, extracts the pixel coordinates of the boundary line in the vertical axis direction in adjacent image frames, analyzes whether the track extension is toward the deep layer area, identifies the track segment with the longitudinal close trend in the continuous frames, screens the positions in the frame sequence, and obtains a longitudinal contact path set. The contour closed relationship comparison submodule obtains the edge trend between the upper boundary and the middle layer path in the corresponding frame image based on the longitudinal contact path set, judges whether the paths form a closed zone in the image, extracts the path number index constituting the closed relationship, traces the connection sequence in the continuous frames, and obtains a structure traction concentrated path sequence.

[0011] As a further scheme of the present application, the tissue boundary disturbance recognition module comprises: The structure trend extraction submodule extracts the edge contour lines on both sides of the structure band in adjacent image frames based on the continuous structure band in the structure traction concentrated path sequence, collects the trend direction data of the contour segment, extracts the extension angle sequence according to the spatial changes on the horizontal axis and the vertical axis, compares the direction trend of the same position in the differentiated frame image, and obtains a structure contour direction sequence. The reverse offset judgment submodule calls the structure contour direction sequence, identifies the continuous image segment with the reverse change in direction in the frame sequence, extracts the trend continuous line segment of the reverse offset in adjacent frames, extracts the image frame number corresponding to the curve segment and the coordinate information of the contour line on the space path, and obtains a direction offset contour segment set. The path number tracking submodule extracts the time position index corresponding to the curve change according to the image frame sequence based on the set of direction offset contour segments, arranges the continuously changing path line segments according to the frame sequence, extracts the path number of the direction continuously offset line segment, and obtains the wound boundary drift disturbance path set.

[0012] As a further scheme of the present application, the adhesion imbalance path labeling module comprises: The direction structure extraction submodule extracts the corresponding path region in the continuous image frame based on the path number in the wound boundary drift disturbance path set, obtains the image texture with direction in the region, calls the gray direction gradient to extract the structure direction in the region, obtains the direction vector result in each frame, and obtains the arrangement direction sequence data; The angle offset detection submodule compares the direction vector of the path region in the continuous frame with the included angle of the sliding trend direction, screens the continuously offset paragraphs in the included angle change track, extracts the number index of the continuously deviated arrangement direction of the sliding direction in the corresponding frame sequence, and obtains the sliding offset path number set; The trajectory path extraction submodule extracts the spatial position of each group of path numbers in the corresponding frame sequence based on the sliding offset path number set, obtains the coordinate point sequence of the same path in the continuous frame, arranges the coordinate positions corresponding to the path in time sequence, and obtains the adhesion structure imbalance path coordinate set.

[0013] As a further scheme of the present application, the wound evaluation management output module comprises: The path region extraction submodule obtains the corresponding region in the image layer based on the path number in the adhesion structure imbalance path coordinate set, extracts the structure texture and boundary morphology change in the image block, obtains the image features of the region covered by the path, and obtains the path covered tissue block set; The tissue performance extraction submodule calls the path covered tissue block set, monitors the exudation and extension track in the block, extracts the contour boundary contact condition, obtains the tension direction change in the texture region, identifies the image region with performance characteristics, and obtains the tissue state performance feature set; The image layer state mapping submodule obtains the performance difference region boundary in the image layer based on the tissue state performance feature set, extracts the matching pixel pattern according to the path region position, maps the state pattern in the image layer, and obtains the wound multi-source state management image layer structure.

[0014] On the other hand, a large-area burn intelligent wound evaluation management method is based on the above-mentioned large-area burn intelligent wound evaluation management system and comprises the following steps: S1: Obtain a burn wound differential image, identify the edge curvature of a high reflection area, combine the texture extension direction, analyze the frame sequence intersection trend, screen the area with consistent and continuous extension direction, extract the corresponding position, and obtain a superficial exudation associated area set; S2: Based on the superficial exudation area set, select the middle layer tissue boundary trend, analyze the longitudinal offset, compare the deep layer approaching direction, extract the image sequence according to the closed state of the path and the upper edge, and obtain a structure traction path sequence; S3: Based on the structure traction path sequence, extract the structure on both sides of the edge direction, compare the frame trend change, track the reverse offset edge segment, and extract the change path frame by frame to obtain a wound boundary disturbance path set; S4: Based on the wound boundary disturbance path set, track the path structure trend, compare the slip angle change, extract the path area in the continuous deviation stage, and obtain a path coordinate set of adhesion structure imbalance; S5: Based on the path number of the imbalance path coordinate set, view the layer coverage area, analyze the exudation expansion, boundary connection and tension direction, distinguish the path state difference, and obtain a wound multi-source state management layer structure.

[0015] Compared with the prior art, the advantages and positive effects of the present application are: In the present application, by extracting the correlation between the exudation area edge trend and the texture extension direction, combining the longitudinal contact trajectory of the middle layer tissue boundary in the image sequence, iteratively identifying the angle change process of the slip trend and the local direction information, tracking the continuous change of the path offset behavior in the time range, extracting the spatial distribution of the structure response characteristics in the layer area, and correlating the tension direction, boundary movement and exudation trend, the layer output of the regional state difference is completed, the spatiotemporal correlation range of the path change and the regional difference in the image sequence is expanded, and the linkage ability of the structure path and the tissue state under the continuous image is enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The system flowchart of the present application; Figure 2 The system block diagram of the present application; Figure 3 The flowchart of the exudation associated area extraction module in the present application; Figure 4Flow chart of the structure stress concentration identification module in the present application; Figure 5 Flow chart of the tissue boundary disturbance identification module in the present application; Figure 6 Flow chart of the adhesion imbalance path calibration module in the present application; Figure 7 Flow chart of the wound evaluation management output module in the present application; Figure 8 Flow chart of the method steps in the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the present application will be described below in combination with the drawings.

[0019] In the embodiments of the present application, the words such as "example", "for example" are used to represent as an example, illustration or description. Any embodiment or design scheme described as "example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. In fact, the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present application, the meaning expressed by "and / or" can be both, or can be one of the two options.

[0020] In the embodiments of the present application, "image" and "picture" can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent. "Of", "corresponding" and "relevant" and "corresponding" can be used interchangeably at times. It should be pointed out that when the distinction is not emphasized, the meanings expressed are consistent.

[0021] In the embodiments of the present application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1. When the distinction is not emphasized, the meanings expressed are consistent.

[0022] In order to make the technical problems, technical solutions and advantages of the present application clearer, the following will be described in detail in combination with the drawings and specific embodiments.

[0023] The embodiments of the present application provide a large-area burn intelligent wound evaluation management system, as shown in the large-area burn intelligent wound evaluation management system schematic diagram, the system comprises: Figures 1-2 ​The exudation-related region extraction module acquires differential layer images of burn wounds, calls the curvature features of the edges of high liquid reflectivity areas in the images and the local texture extension direction, analyzes the continuity of the cross trend direction between the edge trend direction and the main texture direction, extracts exudation texture regions with consistent direction and continuous extension in consecutive image frames, and classifies the corresponding regions into the area set in the image to obtain the shallow exudation-related area set. The structural stress concentration identification module is based on the set of shallow exudation associated areas. It locates the trend of the middle tissue boundary line in the corresponding image region, tracks the longitudinal offset trend of the boundary line in adjacent frame images, and uses the contour path of contact in the deeper direction in consecutive frames as the judgment basis. It compares and numbers the regions where the upper edge is closed to obtain the structural traction concentration path sequence. The tissue boundary disturbance identification module is based on continuous structural bands in the structural traction concentrated path sequence. It compares the directionality of the edge contour curves in the image frames on both sides of the structural band, extracts the continuous frame changes with the opposite trend direction, filters the contour segments with continuous bending change behavior, and labels the change path sequence number according to the frame order to obtain the wound boundary drift disturbance path set. The adhesion imbalance path calibration module is based on the path numbering of the path set of the wound boundary drift disturbance path. It observes the main direction of cell arrangement in the corresponding path area in the continuous image frame sequence and compares the offset angle change with the slip trend direction. The path frame image that shows the slip direction continuously deviating from the arrangement direction is classified into the area where the structure offset trajectory is located according to the number, and the adhesion structure imbalance path coordinate set is obtained. The wound assessment and management output module is based on the path numbering of the path coordinates of the adhesion structure imbalance path. It analyzes the corresponding tissue manifestation type on the layer according to the path coverage area, and associates the expansion trend of exudate, the connectivity of boundary movement and the propagation state characteristics of tension within the path range. Each type of image area is covered by a layer according to the manifestation difference to obtain the multi-source status management layer structure of the wound.

[0024] The superficial exudate associated area set includes a high-reflectivity edge curvature concentration area, a texture main direction continuous area, and a direction intersection consistent area. The structural traction concentration path sequence includes a continuous offset line of the middle tissue boundary, a deep contact contour path, and a closed trend corresponding area. The wound boundary drift disturbance path set includes a reverse offset continuous frame segment, a bending change contour segment, and an edge trend disturbance path. The adhesion structure imbalance path coordinate set includes slip offset trajectory coordinates, an arrangement direction deviation area, and a structural offset aggregation path. The wound multi-source state management layer structure includes a tissue type layer, an exudate expansion trend layer, a boundary connectivity state layer, and a tension propagation distribution layer.

[0025] Specifically, such as Figure 2 , 3 As shown, the exudation-related region extraction module includes: The image hierarchy recognition submodule acquires differentiated images of burn wounds, extracts grayscale boundaries of highly reflective liquid areas, collects edge bands with brightness jumps, extracts grayscale direction and distribution trend of edge pixels, locates image positions between grayscale breakpoints, and obtains a set of reflective edge regions. When acquiring differential images of burn wounds, a tissue moisture distribution feature layer can be extracted from the red channel image, and compared with the epidermal exudate area obtained from the green channel. The contrast differences between different regions within the grayscale range are collected, and the preliminary boundaries of liquid areas are extracted from regions with grayscale abrupt changes greater than 80. Combined with the brightness distribution map of the current image, the edge change area is delineated by comparing boundary pixels with grayscale gradient changes greater than 30. Horizontal and vertical grayscale scans are performed on this area to obtain the edge band where the brightness jump occurs. Then, the grayscale direction of each pixel in the edge band is determined, and the grayscale change rate of the pixel in eight adjacent directions is collected. The direction with the largest change rate is selected as the grayscale extension direction of that pixel. Subsequently, the image near the edge line is processed... The overall trend of the blocks is compared, and the distribution direction of edge line segments is extracted in the same area. It is then identified whether there is a break in the direction change of adjacent line segments. If the angle of abrupt change in grayscale extension direction is greater than 45 degrees, it is determined to be a grayscale breakpoint. The pixel distribution at both ends of the breakpoint is further detected, and pixel interval segments with grayscale value changes greater than 60 are screened out. The corresponding image coordinate range is extracted in these areas and output as the localized area information. Combined with the continuity of edge changes, background stray layers are screened out according to the threshold condition that the area is greater than 1200 pixels. In the remaining areas, the contour is screened according to the degree of boundary line closure. Blocks with an edge break rate greater than 0.3 are removed, and areas that meet the contour closure standard are selected as the effective response areas of the high reflectivity area edges, thus obtaining the set of reflectivity edge areas.

[0026] The edge trend analysis submodule extracts the edge curvature direction and texture extension direction based on the set of reflective edge regions, compares the cross trends in space, compares the trend continuity in the image frame sequence, extracts the frame segment images with consistent cross directions, and obtains a set of continuous cross trend segments. First, the pixel distribution of each edge region is selected. Angle measurements are performed in a single frame image according to the edge cues. The horizontal and vertical coordinates of three adjacent points are substituted into a difference form to extract the curvature direction angle of each point, and the linear change direction is identified. Then, the texture region is extracted by expanding outward from the center of the image patch. A radius is defined outside the edge region, and the gray-level direction of each pixel block within this radius is extracted. The main direction of the gray-level gradient of each point is collected, and the direction of the maximum pixel gray-level difference is taken as the texture extension direction. The angle between the edge curvature direction and the texture extension direction is determined, and edge pixels with an angle change of less than 15 degrees are selected as potential intersection trend starting points. Then, the corresponding positions are tracked along the starting points in adjacent image frames. Pixel segments with the same directional relationship in consecutive image frames are extracted. When judging directional stability, a comparison interval of 3 frames is used to extract the angle difference of the path direction in each frame. If the angle change between consecutive frames is less than 10 degrees, it is considered that the direction is consistent. By judging the continuous direction, the index of the continuous region is extracted on multiple image frames. In actual operation, if a path shows that the angle between it and the initial edge direction is maintained within 10 degrees in the 6th, 7th and 8th frames of the image, then the path can be regarded as a stable continuation of the cross direction. Finally, the image frame index of the region that meets the directional continuity condition is extracted and the number is output. The corresponding image segment is classified into the image segment with consistent cross trend, and the set of cross trend continuous segments is obtained.

[0027] The extended area extraction submodule is based on the set of cross-trend continuous segments. It tracks the texture region in the image frame sequence, filters the positions with consistent directional extension, extracts the covered area in the time series, locates the corresponding map patch, and obtains the set of shallow exudation associated areas. First, extract the texture region corresponding to the starting point of the cross-path in each frame. Establish the range of the coordinate starting point extending outward according to the edge direction. Sequentially extract the continuous grayscale texture segments under this region in the frame sequence image. In the sample images with frame numbers 4, 5, and 6, extract the 16×16 pixel block to the right of the center of the tile where the same path starts. Monitor the change of the grayscale texture direction offset in the vertical axis. If the offset angle change does not exceed 10 degrees in consecutive frames and the grayscale direction deflection axis points to the upper right direction three times consecutively, then the texture region is determined to be a region with consistent direction. Then continue to track the direction path of the texture extension in this region, and find the next corresponding texture on the same path. The corresponding texture region is extracted from the image frame. In frame number 8, the texture region corresponding to this path is located 12 pixels to the right of the initial region. The texture direction in this region is identified as being consistent with the direction of the previous frame. The consistent position is filtered by judging the degree of alignment of the gray-scale direction of the region. When the consistent direction is interrupted or the angle shift exceeds 20 degrees, the tracking stops. During the entire tracking process, the image position segments between the starting region and the ending region are extracted, and the tile coordinates in these position sequences are spatially located. The image blocks corresponding to the positions of the regions that satisfy the directional extension relationship in the continuous frames are taken as target tiles. The tile coordinates in each region path are labeled according to the row and column range to obtain the set of shallow exudation associated areas.

[0028] Specifically, such as Figure 2 , 4 As shown, the structural stress concentration identification module includes: The boundary line extraction and processing submodule obtains the corresponding position of the middle tissue based on the set of shallow exudation associated areas, extracts the edge band direction of the gray-scale abrupt change area and the edge pixel sequence extending along the horizontal axis, tracks the continuous edge direction in the frame sequence, locates the corresponding trajectory line position in each frame image, and obtains the middle boundary line trajectory set. First, extract the grayscale variation band within the horizontal extension range of the center of each image patch. Then, identify the region with the largest grayscale gradient change in each column of pixel values. In the sequence of frame numbers 6, 7, and 8, select a 12-pixel range extending horizontally to the left from the upper right corner of each region and iterate through the grayscale values. Pixels with a grayscale difference greater than a set threshold are designated as grayscale abrupt change areas. Extract the edge path based on the coordinate range between the upper and lower boundaries of these grayscale abrupt change areas. Calculate the horizontal extension length of the grayscale abrupt change path in each frame. Filter image segments with consecutive edge paths exceeding 10 pixels, sort them by frame number, and extract the spatial trajectory lines of the edge paths of these image segments in different frames. The pixel coordinate sequence appearing in each frame on the trajectory line is numbered according to its arrangement order in the horizontal axis of the image, forming the intra-frame edge pixel path. Then, the intra-frame paths are spliced ​​in ascending order of frame number. The overall offset trend of the path direction in the vertical axis of consecutive frames is analyzed, and the coordinate points at the path interruption are corrected. In the example of processing image sequence number 10 to 20, edge bands with an extension length of more than 15 pixels in the horizontal axis are extracted, and their vertical axis starting coordinate positions are mapped frame by frame. The coordinate path between the starting position and the ending position is marked in each frame. The continuous splicing processing of all edge paths in the time series is completed to obtain the middle layer boundary line trajectory set.

[0029] The longitudinal path change recognition submodule calls the middle layer boundary line trajectory set, extracts the pixel coordinates of the boundary line in the longitudinal direction in adjacent image frames, analyzes whether the trajectory extension is towards the deep region, identifies trajectory segments with longitudinal approach trend in continuous frames, filters the positions in the frame sequence, and obtains the longitudinal contact path set. First, the positional change of the trajectory line along the vertical axis is extracted. The coordinate difference of the path along the vertical axis in consecutive frames is obtained. For frames numbered 10 to 25, the bottom pixel position on the vertical axis is extracted sequentially. The monotonicity of the vertical axis coordinates between adjacent frames is calculated. Sequences maintaining downward displacement are grouped by number and arranged in ascending order according to frame number. The intervals of continuous vertical change in each path segment are marked. Further, segments in the frame segment whose vertical coordinates decrease by more than 5 pixels compared to the previous frame are identified. Segments with a continuous decrease of 3 or more frames are defined as vertically shifting segments. For effective trajectory segments with near-trends, when performing vertical trend judgment operations on the sequence segments of each trajectory, the change range of the median coordinate in the vertical axis direction is used as the main reference. If the vertical coordinate of the median line in a certain frame continuously decreases and the change range exceeds 20 pixels, and there is no obvious break or reversal in the trajectory, it is classified as a continuous trend segment. Further, the distance between the vertical coordinate of the deepest point of the path line in the last frame of each segment and the pixel distance between the image bottom edge is judged. When the distance value is less than the set lower limit threshold of 25 pixels, it is classified into the sequence segment that is moving towards the deeper layer. Then, the path segments are numbered and classified according to the frame sequence information to obtain the set of vertical contact paths.

[0030] The contour closure relationship comparison submodule is based on the set of longitudinal contact paths. It obtains the edge direction between the upper boundary and the middle layer path in the corresponding frame image, determines whether the paths form a closed zone in the image, extracts the path number index that constitutes the closure relationship, and tracks the connection sequence in consecutive frames to obtain the path sequence in the structural traction set. First, based on the spatial location of the path in the frame, the edge pixel positions and gray-level abrupt changes of each path in the corresponding region of the upper boundary of the image are extracted. The intersection region of the path's upper edge extension direction and the upper boundary line is obtained. It is determined whether the two paths form a closed curve extending downward from the upper boundary to the middle layer path in the image. The pixel distribution boundaries in the extracted regions are screened to see if they enclose a closed region. In the judgment process, whether the path's start point and end point are located in the upper left and lower right quadrants of the image is used as one of the reference standards. At the same time, whether the gray-level difference at the horizontal pixel connection point of the closed shape edge is less than 30 is used as the continuity of connection. Path pairs are further extracted from the closed region and numbered as P14 and P2. For path 1, extract the number of connected points forming closed regions in each frame of the images with frame sequence numbers F12 to F20. When the proportion of closed point connections in consecutive frames is greater than 70% of the path boundary length, it is determined that the path pair forms a closed relationship in that frame. Record the closure status of the path pair in consecutive frames. In frame segments with a large number of path connections, such as F13 to F18, the number of closed regions formed by path numbers P14 and P21 reaches more than 3. Then, perform a number tracking operation on the closed path set, matching whether the path appears continuously in the image frame sequence and forms a connection frame by frame. Sort the path segments corresponding to the number order into a connection sequence according to the frame order to obtain the path sequence in the structural traction set.

[0031] Specifically, such as Figure 2 , 5 As shown, the organization boundary disturbance identification module includes: The structure orientation extraction submodule extracts the edge contour lines on both sides of the structure band in the structure traction concentrated path sequence based on the continuous structure band. It collects the trend direction data of the contour segment, extracts the extension angle sequence according to the spatial changes on the horizontal and vertical axes, and compares the direction of the same position in the differentiated frame images to obtain the structure contour direction sequence. First, the regions where the edges of the path band are located on both sides are extracted according to the frame number identified in the image frame sequence. The pixel positions of the contour line located at the structural boundary are collected from these regions. 50 pixels are extracted from the left and right sides of each path pair. The spatial displacement between adjacent points is calculated along the horizontal and vertical axes to obtain the spatial direction of a single contour curve. The current curve orientation is determined according to the displacement relationship of adjacent points on the two axes. The direction change is obtained by comparing the trend difference of the contour line at the same position in different frame images. Images with frame numbers F22 to F26 are selected, and the contour direction change angle values ​​on both sides of path number S9 are extracted. Frame segments with a change angle within 15 degrees are counted as direction-maintaining segments. If the contour angle change on the same side in consecutive frames is greater than 45 degrees, the segment is determined to be a direction-changing segment. The direction data sequence is further extracted from the continuous contour segments. A contour extension direction list is established according to the horizontal and vertical axes, respectively. The increase or decrease relationship between the change direction of each path segment in the frame sequence and the previous frame is recorded. Finally, the extension angle trend of the contour segments in space is summarized and arranged to obtain the structural contour direction sequence.

[0032] The reverse offset judgment submodule calls the structural contour direction sequence to identify continuous image segments whose direction changes in the reverse order in the frame sequence, extracts the trend of the reverse offset in adjacent frames, extracts the image frame number corresponding to the curve segment and the coordinate information of the contour line on the spatial path, and obtains a set of direction offset contour segments. First, the directional trend value of each numbered path in the time sequence is extracted from the frame image sequence. The directional trend of the same path in consecutive frames is compared. In frames F33 to F36, the directional change of path P7 is observed. When the direction is 45 degrees in frames F33 and F34, changes to a negative direction of 15 degrees in F35, and a negative direction of 42 degrees in F36, it can be determined that the path direction shows a reverse trend. Further, paths with the same reverse change pattern are screened frame by frame in the structural contour direction sequence. A path segment list is constructed according to the number identifier. The pixel positions of the horizontal and vertical axes in each segment are extracted. Combined with the image coordinate system, the paths are... The trajectory is converted into a spatial trajectory segment. It is then determined whether the trajectory remains continuous in two consecutive frames. If the trajectory does not break in frames F35 to F38 and the direction remains reversed, it is included in the valid reverse segment. Subsequently, the frame number of each trajectory segment is extracted from the direction data and used as a time axis identifier to record the position change of the segment. The point set sequence that constitutes the curve is extracted from each segment and mapped to the image space. The start and end coordinates of the contour line are marked to form a dataset containing the start point, end point, direction attribute and corresponding frame number. Finally, continuous segments of the reverse contour curve in the image sequence are extracted from all reverse change trajectory segments that meet the conditions and numbered to obtain a set of direction offset contour segments.

[0033] The path number tracking submodule is based on the set of directional offset contour segments. It extracts the time position index corresponding to the curve change according to the image frame sequence, arranges the continuously changing path segments in frame order, extracts the path number of the continuously offset line segments, and obtains the wound boundary drift disturbance path set. First, the coordinate information of the curve segment is read in each image frame. A sequential sequence is established according to the image frame number, and the path number is bound to the frame number in pairs. The time position index of each path segment is confirmed by the frame sequence number axis. For example, if the path number P12 appears sequentially in frames F21 to F27, then P12 is bound to the frame interval F21 to F27 and identified as the sequence interval L12. At the same time, the spatial coordinate changes of the path segment in each frame image are read, and the curvature trend is checked for continuous angular shifts. If the angle is 22 degrees in frame F21, 27 degrees in F22, and rises to 34 degrees in F23, and the angle change direction is consistent for three consecutive frames, it is included in the continuous change state. If a sudden angle change occurs in the same path segment... If the direction is opposite, the sequence extraction is terminated to ensure that the extracted path segments have a continuous directional offset feature. The number of the continuously offset path segments is recorded. By traversing all frames in the sequence that meet the feature, a set of numbered sequences is established. When path numbers P15, P16, and P19 all meet the continuous offset requirement and the displacement amplitude between the start and end points is greater than the set minimum value of 5 pixels, they are included in the valid number set. The displacement records and angle offset records of all numbered paths in the image frames are further extracted from them, and a time series table of corresponding frame numbers is formed. All numbered paths are classified and organized according to frame numbers. Finally, the path numbers with consistent directions are extracted and the number sequence is output to obtain the wound boundary drift disturbance path set.

[0034] Specifically, such as Figure 2 , 6 As shown, the adhesion imbalance path calibration module includes: The orientation structure extraction submodule extracts the corresponding path region in consecutive image frames based on the path number in the path set of the wound boundary drift disturbance path, obtains the directional image texture in the region, calls the gray-scale orientation gradient to extract the structural orientation in the region, obtains the orientation vector result in each frame, and obtains the arrangement orientation sequence data. First, the position index of each path in the image frame sequence is obtained according to the numbering order. The image region corresponding to the number is extracted frame by frame. Coordinate blocks are cropped from each frame image by calling the image reading method. The corresponding coordinate information is obtained from the path offset record of the previous stage. The image size is set to a square window with a width and height of 50 pixels. Zero-padding is performed on the edges to fill in empty areas. After completing the cropping of the original region blocks, texture elements with directional features are identified at the pixel level for each image region. Texture variation differences are extracted by traversing the grayscale values ​​of pixels and constructing a 3×3 neighborhood. The grayscale difference between the center pixel and its neighboring pixels in the horizontal and vertical directions is calculated. The difference direction is used as the basic index to record the initial gradient direction. Then, the texture variation trend is superimposed layer by layer through continuous row and column sampling to extract the cumulative change in the gradient direction. Furthermore, a filtering operation is performed on the gray-level gradient values. When the cumulative gray-level difference in a certain direction is greater than the set gray-level change limit value Δg, that direction is regarded as the main structural direction of the region. Δg is set to 14 based on the value results of typical exudation area images in the training samples. Values ​​lower than this value are not recorded. Combining the sample of region number 23 in a certain frame of the example image, the gray level of the center point of the gray-level window is 98, the gray levels of the adjacent upper and lower pixels are 112 and 105, and the horizontal gray levels are 88 and 91, respectively. The vertical difference is calculated to be 14, and the horizontal difference is 13. Only the vertical difference is greater than Δg. Therefore, the local main structural direction of the window is determined to be vertical. By repeating the above calculation frame by frame, the direction vector is extracted in all image frames. The direction vector values ​​obtained in each frame are sequentially numbered and recorded to obtain the arranged direction sequence data.

[0035] The angle offset detection submodule calls the arrangement direction sequence data, compares the angle between the direction vector of the path region in the continuous frames and the sliding trend direction, filters the continuous offset segments in the angle change trajectory, extracts the number index of the continuous deviation of the sliding direction from the arrangement direction in the corresponding frame sequence, and obtains the sliding offset path number set. First, the direction vector values ​​of the corresponding regions for each path in consecutive image frames are extracted. The path numbers are aligned with the frame order to construct a two-dimensional direction sequence matrix. Each column represents the direction vector angle of each path in a certain frame, and each row represents the trajectory of the direction angle change of a certain path. Then, the slippage trend direction is obtained. This direction is derived from the extraction results of the wound boundary line trajectory change in the previous stage, combined with the inter-frame coordinate displacement difference in the same region. By comparing the difference in the vertical and horizontal coordinates of adjacent path segments between two frames in the image coordinate space, the slippage direction angle is calculated based on the arctangent. Then, the angle between the direction vector and the slippage direction is calculated to determine the trend of directional difference. The standard judgment value θth is introduced into the angle calculation as the offset judgment criterion. According to the data, θth is set to 20 degrees. Based on the angle records of consecutive frames with path number 12 in the experimental image, the angles are 83 degrees, 85 degrees, and 87 degrees, corresponding to sliding direction angles of 60 degrees, 62 degrees, and 64 degrees, with included angles of 23 degrees, 23 degrees, and 23 degrees respectively. All of these are greater than θth, indicating that the path has a continuous offset trend. The path number and the corresponding frame sequence index are recorded. In the sequence, further screening is performed on segments in consecutive frames where the included angle change is not less than θth. Only segments with at least 3 consecutive frames are considered as continuous offset segments. These segments are weighted and labeled in the numbered path set, and a dictionary table corresponding to frame number and path number is constructed. All path numbers that meet the condition of continuous deviation of the sliding direction from the arrangement direction are output, resulting in the sliding offset path number set.

[0036] The trajectory path extraction submodule extracts the spatial position of each path number in the corresponding frame sequence based on the slip offset path number set, obtains the coordinate point sequence of the same path in consecutive frames, arranges the coordinate positions corresponding to the path in time order, and obtains the coordinate set of the adhesion structure imbalance path. First, establish a mapping relationship between path number and frame sequence. Read the contour edge of the region identified by the path number from each frame image, and select the midpoint of the edge as the representative coordinate point. If the path number is 08 and the frame sequence number is f1 to f6, then the center point position of path 08 in pixel space needs to be obtained in images f1 to f6 respectively. For example, in f1, this point is (132, 245), and in f2 it is (135, 248), and so on, constructing the coordinate sequence of path number 08. Then, perform the same extraction process for each path number, obtaining the corresponding coordinate points of frame sequences 09, 10, etc., forming a triple matching relationship of path-frame number-coordinate. Further confirm the data correctness by observing the trend of coordinate point changes on the horizontal and vertical axes of the image. If the path jumps between frames, it is necessary to determine whether the failure to extract coordinate points is due to image occlusion. For example, if the f4 point is missing, the coordinates of the missing point are inferred by linear interpolation. Then, the coordinate point sequence corresponding to each numbered path is combined into a time series coordinate vector set according to the frame order. The time series trajectory of each path segment is organized into a continuous trajectory line with the frame number as the horizontal axis and the coordinate position as the vertical axis. All node coordinates of each trajectory line are stored in the coordinate set. If the numbered path spans 10 consecutive frames, the coordinate points of each frame form a broken line trajectory. After organizing all numbered paths in this way, a complete trajectory set is constructed. The continuous coordinate records of all numbered paths in the frame time axis are output to obtain the coordinate set of the adhesive structure imbalance path.

[0037] Specifically, such as Figure 2 , 7 As shown, the wound assessment and management output module includes: The path region extraction submodule obtains the corresponding region in the layer image based on the path number in the set of path coordinates of the adhesive structure imbalance path, extracts the structural texture and boundary morphology changes in the tile, obtains the image features of the area covered by the path, and obtains the path coverage organization tile set. First, when obtaining the corresponding region in the layer image, the mapping record between the path number and the layer image must be called first. The actual coordinate position of each path number in the image frame is located one by one. Based on the spatial coordinate point sequence of each extracted path, the local image block covered by the path in each frame is cropped from the original image. Let the path number be 06, and the path coordinate point in frame image f3 be (120, 236). A rectangular area with a length and width of 40 pixels is constructed with this point as the center. This block is extracted as the path region image. Then, the block is cropped sequentially for the coordinate position of path 06 in all frames to obtain a set of local image regions corresponding to the number 06. Then, the structure texture extraction operation is performed on each block one by one. In the extraction operation, the gray-level change direction of each pixel in the block needs to be counted, and the degree of texture isotropy in the local neighborhood is calculated. A window size of 5×5 is used. The neighborhood is scanned by tiles, and the direction angle of the strongest gray value change in each window is recorded. If the gray value rises continuously in the center of the tile and the angle is concentrated between 30° and 45°, the texture of the area is determined to be directional. Then, the contour is extracted according to the edge shape of the tile. By detecting the position of the gray value change point at the edge of the tile, the boundary contour curve is constructed. If the edge contour bends continuously in the lower left direction, it is recorded as a structural deformation feature. At the same time, the texture density and boundary closure degree in each tile need to be evaluated. By traversing all the numbers in the path number set, the tile extraction, texture acquisition and boundary analysis operations are performed in sequence. All tile image features are recorded according to the path number and bound to the original coordinate position to form a data structure. Each tile is finally stored in the set in the form of a triplet of path number, frame number and image feature, resulting in a path coverage organized tile set.

[0038] The tissue performance extraction submodule calls the path to cover the set of tissue tiles, monitors the seepage extension trajectory in the tiles, extracts the contact of the contour boundary, obtains the tension direction change in the texture region, identifies the image region with performance characteristics, and obtains the tissue state performance feature set. First, obtain the path number corresponding to each tile and its coordinate range in the image. Then, sequentially call the tile data of each frame according to the time order of the tile set. Set the tile number as P12, its corresponding image frame as f6, and its coverage area as a rectangular window (x1=105, y1=220, x2=135, y2=250). Extract the grayscale value of each pixel in the tile and trace the chain of connected pixels changing from low grayscale to high grayscale in that tile. Determine whether the chain shows a spatial extension trend in adjacent frames. If the end of a chain in tile f6 extends to a new high grayscale point in tile f7, record this chain as an effusive extension trajectory. Then, obtain the spatial relationship between the distribution positions of the edge contour points in the tile and the coordinates of the effusive chain ends. If the distance between the end point and the edge curve in the y-axis direction is less than or equal to 3 pixels... If more than 8 pixels overlap along the x-axis, it is determined that contour boundary contact has occurred. A 5×5 pixel sliding window is then constructed within the tile area. The directional tension of the pixel grayscale distribution within each window is determined. The direction of the line connecting the pixel with the maximum grayscale value and the pixel with the minimum grayscale value is taken as the local tension direction. The change in the tension direction angle of consecutive tiles under the same path number is compared. If the angle change value is continuously greater than 30 degrees in three consecutive frames, the tile is marked as a region with significant change in tension direction. By traversing all path numbers in the tile set and performing the above steps on all tiles under each number, the tiles with any two or more of the following features are finally counted: exudation trajectory extension, boundary contact, and change in tension direction. Their numbers, frame indices, and spatial locations are recorded to obtain the tissue state performance feature set.

[0039] The layer state mapping submodule obtains the boundary of the region of difference in appearance in the layer based on the tissue state performance feature set, extracts the matching pixel style according to the path region position, and maps the state style in the layer to obtain the multi-source state management layer structure of the wound. First, extract the feature region number marked in each image patch and its coordinate information in the corresponding layer image. Let the number be F03, the corresponding layer image be L2, and its boundary points be (x1=110, y1=95, x2=135, y2=120). After cropping the bounding box, a candidate patch region is formed. Within this patch region, extract the RGB and grayscale values ​​of each pixel and combine them to form a pixel style vector. Record the style vectors extracted from all feature regions into a style index table according to the patch number. Then, find the corresponding patch region according to the spatial range of each path number in the layer. If the current path number is P25, the corresponding layer is L2, and its coverage area is (x1=112, y1=98, x2=132, y2=118), then the pixel style vector within this range is... The pixel value difference is compared with the style vector in F03. The sum of the RGB difference of each corresponding pixel is calculated and the average value is taken. If the average difference is between [25, 60], the area of ​​the image is marked as an area with a difference in state performance. Then, the style type label corresponding to F03 is selected, and the color mark bound to the label of F03 is filled in the area corresponding to P25 in the layer. For example, if the label bound to F03 is exudation tension style and the corresponding color is red transparent mask, then the P25 area is mapped to the red mask layer block. The same processing is repeated for all performance feature blocks. Corresponding style templates are constructed for different feature types. The mapping results are filled in the layer area by area. Finally, a composite image with style mark superimposed display is generated on the original layer to obtain the multi-source state management layer structure of the wound.

[0040] Please see Figure 8 The aforementioned intelligent wound assessment and management method for large-area burns is based on the above-mentioned intelligent wound assessment and management system for large-area burns, and includes the following steps: S1: Obtain differential images of burn wounds, identify the edge curvature of high reflectivity areas, combine with the texture extension direction, analyze the frame sequence intersection trend, filter areas with consistent direction and continuous extension, extract the corresponding positions, and obtain a set of superficial exudation associated areas. S2: Based on the set of shallow exudation areas, select the trend of the middle layer tissue boundary, analyze the longitudinal offset, compare the direction of approaching the deeper layer, and extract the image sequence according to the closure state of the path and the upper edge to obtain the structural traction concentrated path sequence. S3: Based on the structural traction path sequence, extract the direction of the edges on both sides of the structural band, compare the trend changes between frames, track the reverse offset edge segments, extract the change path frame by frame, and obtain the wound boundary disturbance path set; S4: Based on the set of disturbance paths at the wound boundary, track the trend of the path structure, compare the changes in the slip angle, extract the path area of ​​the continuous deviation stage, and obtain the coordinate set of the path of adhesion structure imbalance. S5: Based on the path number of the unbalanced path coordinate set, view the layer coverage area, analyze exudation expansion, boundary connection and tension direction, distinguish path state differences, and obtain the multi-source state management layer structure of the wound.

[0041] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart wound assessment and management system for large-area burns, characterized in that, The system includes: The exudation-associated region extraction module acquires differential layer images of burn wounds, and extracts continuous exudation texture regions based on the edge curvature characteristics and texture extension direction of high liquid reflectivity areas to obtain a set of superficial exudation-associated regions. The structural stress concentration identification module, based on the set of shallow exudation associated areas, locates the trend of the middle tissue boundary in the corresponding image region, tracks the longitudinal offset path and deep contact contour in consecutive frames, and obtains the structural traction concentration path sequence. The tissue boundary disturbance identification module, based on the continuous structural bands in the structure traction concentrated path sequence, compares the contour directional changes in the image frames on both sides, extracts the frame segments with reverse trend shift, and obtains the wound boundary drift disturbance path set; The adhesion imbalance path calibration module, based on the set of path deviations and disturbances at the wound boundary, observes the offset angle between the main direction of cell arrangement and the slippage trend in consecutive frames, identifies the imbalance offset trajectory, and obtains the coordinate set of the adhesion structure imbalance path. The wound assessment and management output module analyzes the tissue performance, exudation trend and tension status within the path coverage area based on the coordinate set of the adhesion structure imbalance path, and obtains the multi-source status management layer structure of the wound.

2. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The set of superficial exudate-related areas includes a high-reflectivity edge curvature concentration area, a texture main direction continuous area, and a direction intersection consistent area. The structure traction concentration path sequence includes a continuous offset line of the middle tissue boundary, a deep contact contour path, and a closed trend corresponding area. The set of wound boundary drift disturbance paths includes a reverse offset continuous frame segment, a bending change contour segment, and an edge trend disturbance path. The set of adhesion structure imbalance path coordinates includes slip offset trajectory coordinates, an arrangement direction deviation area, and a structure offset aggregation path. The wound multi-source state management layer structure includes a tissue type layer, an exudate expansion trend layer, a boundary connectivity state layer, and a tension propagation distribution layer.

3. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The continuous exudation texture region refers to the region that is extracted from the liquid reflective edge of the image, and then tracked frame by frame along the texture direction to select regions with consistent and continuous direction in multiple frames, and to identify segments with continuous temporal and spatial trends. The deep contact contour refers to the longitudinal movement in the frame sequence after extracting the middle layer boundary, and the identification path continuously moving deeper and forming an enclosing trend with the upper layer edge.

4. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The contour directional change refers to tracking the direction of tissue edges in continuous images, analyzing whether there is a reverse offset and abrupt change in trend in adjacent frames, and whether there is a continuous change in time. The imbalance offset trajectory refers to the arrangement direction and sliding direction of the extracted path region. The analysis of each frame shows whether there is a long-term inconsistency. When the deviation phenomenon continues to occur and is accompanied by the spatial displacement of the path in the image, it can be judged as an imbalance offset trajectory.

5. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The exudation-related region extraction module includes: The image hierarchy recognition submodule acquires differentiated images of burn wounds, extracts grayscale boundaries of highly reflective liquid areas, collects edge bands with brightness jumps, extracts grayscale direction and distribution trend of edge pixels, locates image positions between grayscale breakpoints, and obtains a set of reflective edge regions. The edge trend analysis submodule extracts the edge curvature direction and texture extension direction based on the set of reflective edge regions, compares the cross trends in space, compares the trend continuity in the image frame sequence, extracts the frame segment images with consistent cross directions, and obtains a set of continuous cross trend segments. The extended area extraction submodule, based on the set of cross-trend continuous segments, tracks the texture regions in the image frame sequence, filters positions with consistent directional extension, extracts the coverage areas in the time series, locates the corresponding patches, and obtains a set of shallow exudation associated areas.

6. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The structural stress concentration identification module includes: The boundary line extraction and processing submodule obtains the corresponding position of the middle tissue based on the set of shallow exudation associated areas, extracts the edge band direction of the gray-scale abrupt change area and the edge pixel sequence extending along the horizontal axis, tracks the continuous edge direction in the frame sequence, locates the corresponding trajectory line position in each frame image, and obtains the middle boundary line trajectory set. The longitudinal path change recognition submodule calls the middle-layer boundary line trajectory set, extracts the pixel coordinates of the boundary line in the longitudinal direction in adjacent image frames, analyzes whether the trajectory extension is towards the deep region, identifies trajectory segments with longitudinal approach trend in continuous frames, filters the positions in the frame sequence, and obtains the longitudinal contact path set. The contour closure relationship comparison submodule, based on the set of longitudinal contact paths, obtains the edge direction between the upper boundary and the middle layer path in the corresponding frame image, determines whether the paths form a closed zone around each other in the image, extracts the path number index that constitutes the closure relationship, tracks the connection sequence in consecutive frames, and obtains the path sequence in the structural traction set.

7. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The tissue boundary disturbance identification module includes: The structure orientation extraction submodule extracts the edge contour lines on both sides of the structure band in the continuous structure band in the structure traction concentrated path sequence, collects the trend direction data of the contour segment, extracts the extension angle sequence according to the spatial changes on the horizontal and vertical axes, and compares the direction of the same position in the differentiated frame images to obtain the structure contour direction sequence. The reverse offset judgment submodule calls the structure contour direction sequence to identify continuous image segments whose direction changes in the reverse order in the frame sequence, extracts the trend of the reverse offset in adjacent frames, extracts the image frame number corresponding to the curve segment and the coordinate information of the contour line on the spatial path, and obtains a set of direction offset contour segments. The path number tracking submodule, based on the set of directional offset contour segments, extracts the time position index corresponding to the curve change according to the image frame sequence, arranges the continuously changing path segments in frame order, extracts the path number of the continuously offset line segments, and obtains the wound boundary drift disturbance path set.

8. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The adhesion imbalance path calibration module includes: The orientation structure extraction submodule extracts the corresponding path region in consecutive image frames based on the path number in the set of path drift disturbances at the wound boundary. It then obtains the directional image texture within the region, calls the gray-scale directional gradient to extract the structural orientation within the region, and obtains the orientation vector result in each frame to get the arrangement orientation sequence data. The angle offset detection submodule calls the arrangement direction sequence data, compares the angle between the direction vector of the path region in the continuous frames and the sliding trend direction, filters the continuous offset segments in the angle change trajectory, extracts the number index of the continuous deviation of the sliding direction from the arrangement direction in the corresponding frame sequence, and obtains the sliding offset path number set. The trajectory path extraction submodule extracts the spatial position of each path number in the corresponding frame sequence based on the slip offset path number set, obtains the coordinate point sequence of the same path in consecutive frames, arranges the coordinate positions corresponding to the path in chronological order, and obtains the coordinate set of the adhesion structure imbalance path.

9. The intelligent wound assessment and management system for large-area burns according to claim 1, characterized in that, The wound assessment and management output module includes: The path region extraction submodule obtains the corresponding region in the layer image based on the path number in the set of coordinates of the unbalanced adhesive structure path, extracts the structural texture and boundary morphology changes in the tile, obtains the image features of the area covered by the path, and obtains a set of path-covered organization tiles. The tissue performance extraction submodule calls the path-covered tissue patch set, monitors the seepage extension trajectory in the patch, extracts the contact situation of the contour boundary, obtains the tension direction change in the texture region, identifies the image region with performance characteristics, and obtains the tissue state performance feature set. The layer state mapping submodule obtains the boundary of the region of difference in appearance in the layer based on the set of tissue state performance features, extracts the matching pixel style according to the path region position, maps the state style in the layer, and obtains the multi-source state management layer structure of the wound.

10. A method for intelligent assessment and management of large-area burn wounds, characterized in that, The execution of the intelligent wound assessment and management system for large-area burns according to any one of claims 1-9 includes the following steps: S1: Obtain differential images of burn wounds, identify the edge curvature of high reflectivity areas, combine with the texture extension direction, analyze the frame sequence intersection trend, filter areas with consistent direction and continuous extension, extract the corresponding positions, and obtain a set of superficial exudation associated areas. S2: Based on the set of shallow exudation areas, select the trend of the middle layer tissue boundary, analyze the longitudinal offset, compare the direction of approaching the deeper layer, and extract the image sequence according to the closure state of the path and the upper edge to obtain the structural traction concentrated path sequence. S3: Based on the structural traction path sequence, extract the edge directions on both sides of the structural band, compare the trend changes between frames, track the reverse offset edge segments, extract the change path frame by frame, and obtain the wound boundary disturbance path set; S4: Based on the set of disturbance paths at the wound boundary, track the trend of the path structure, compare the changes in the slip angle, extract the path region of the continuous deviation stage, and obtain the set of coordinates of the path of adhesion structure imbalance. S5: Based on the path number of the imbalance path coordinate set, view the layer coverage area, analyze the exudation expansion, boundary connection and tension direction, distinguish the path state differences, and obtain the multi-source state management layer structure of the wound.

Images