Automatic burn area evaluation method based on AI image recognition

Abstract

The present application relates to the technical field of medical image analysis, in particular to a burn area automatic evaluation method based on AI image recognition, comprising the following steps: analyzing image gradient sequence, screening for inflection points and encoding texture order, analyzing brightness difference to obtain attenuation trend to generate degradation structure, comparing tension mutation and convergence direction, screening for bridging points, identifying color difference density inversion to construct boundary point set, analyzing curvature difference to perform area expansion correction, in the present application, the texture order identifier is constructed by analyzing the pixel gradient sequence in the original image, the regional damage intensity judgment mechanism is established by combining the brightness change and the thermal damage signal difference, the wound edge point structure is derived by the tension direction trend, the continuous boundary range is identified by combining the color difference density, the area expansion comparison and correction are carried out by fusing the curvature difference, the systematic evaluation logic from boundary recognition to area mapping is established, the damage range is accurately delimited and the body surface characteristics are matched, and the precision evaluation and regional expansion of the burn area are realized.

Description

Technical Field

[0001] This invention relates to the field of medical image analysis technology, and in particular to an automatic burn area assessment method based on AI image recognition. Background Technology

[0002] The field of medical image analysis technology refers to the technical direction of using computer algorithms to deeply mine and quantitatively evaluate medical image data. Specifically, it includes clinical image enhancement, lesion segmentation, feature extraction, anatomical structure recognition, and quantitative measurement. By constructing digital pathological models, computers can simulate doctors' interpretation of human tissue morphology, lesion severity, and spatial topological relationships. This enables systematic image information processing in areas such as auxiliary diagnosis, efficacy evaluation, lesion localization, and contour determination. Typically, it relies on a large number of labeled medical samples to build a training set, and uses structures such as deep neural networks to achieve pixel-level or region-level classification, allowing computers to extract visual representations of different pathological tissues or different injury categories from raw medical imaging images. Differences support the realization of various clinical quantitative analysis tasks. Among them, the traditional burn area assessment method refers to the process in which medical staff identify the burn wound boundary by visual inspection based on the wound images taken on site or by directly observing the patient's skin condition. They also use methods such as drawing a body surface map using the nine-point method, comparing skin areas with a ruler, calculating the local and overall proportions using the human body surface area formula, or estimating the damaged area by applying a transparent mesh film to the skin surface and measuring the damaged area grid by grid. The method relies on manual observation of burn color changes, eschar extent, and the boundary with healthy skin, and manual calculation based on experience or standard body surface maps to complete the qualitative identification and semi-quantitative estimation of the burn area.

[0003] Traditional burn area assessment techniques rely on a combination of visual inspection and physical measurement to evaluate burn area. This method depends on the subjective judgment of medical staff regarding the color, texture, and extent of the burn wound. It also uses ruler comparison, surface map reference, or transparent grid to assist in area estimation. This method is highly dependent on the operator's clinical experience. In medical photography, the blurred wound edges of burn tissue and the color gradation caused by inflammatory reactions often lead to segmentation errors. In the process of quantifying the area, there are problems such as unclear division of irregular boundaries and inaccurate estimation scale. It is difficult to correct for geometric projection errors caused by the curvature of human skin during manual recording and conversion of body surface proportions. As a result, the area assessment results fluctuate greatly with changes in personnel, experience, and measuring tools, and cannot achieve the high consistency and standardization required for precision medicine. Summary of the Invention

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an automatic burn area assessment method based on AI image recognition, comprising the following steps: S1: Call the original image of the burn area, analyze the gradient changes, sort the gradient sequence in the pixel neighborhood, compare the changes of adjacent gradient terms, filter inflection points and establish jump markers, combine them with the gradient sequence positions to form a jump rate distribution and perform mapping to form texture order encoding and generate damage texture order parameters. S2: Based on the damage texture order parameters, analyze the brightness changes in the corresponding area, evaluate the attenuation trend by calculating the difference between brightness and thermal loss signal, and take the average of the attenuation trend in multiple directions to determine the attenuation intensity and establish a degradation marker to generate degradation intensity structural parameters. S3: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the wound edge region, screen abrupt change points and establish breakpoint markers by comparing tension differences, perform convergence judgment on the tension directions on both sides of the breakpoint, deduce the bridging direction in the convergence region and screen stable points, and generate wound edge bridging association parameters. S4: Based on the aforementioned wound bridging correlation parameters, analyze the color difference gradient density change in the corresponding region, identify the reversal point and establish pulse markers by comparing the enhancement and attenuation directions of the density sequence, sort the pulse markers and filter key points, construct a continuous boundary point set, and generate the burn location identification boundary. S5: Identify the boundary based on the burn location, analyze the curvature change of the corresponding skin area in the orthogonal direction, determine the dominant and secondary directions by comparing the curvature differences, perform stretching adjustment on the area projection in the dominant direction, perform contraction adjustment on the secondary direction, organize the adjusted area mapping into structural output, and generate area stretching correction results.

[0005] As a further aspect of the present invention, the damage texture order parameters include gradient energy level jump distribution coefficient, texture order level encoding, and regional gradient sequence feature quantity; the degradation intensity structure parameters include local brightness decay curve index, thermal damage progression level marker, and degradation structure partition index; the wound edge bridging correlation parameters include tension breakpoint correlation coefficient, cross-segment bridging path index, and wound edge connectivity stability measure; the burn location identification boundary includes color difference density pulse feature quantity, boundary key point coordinate set, and wound edge contour continuity index; and the area stretching correction results include orthogonal curvature direction ratio, principal and secondary direction area stretching factor, and body surface area mapping correction coefficient.

[0006] As a further aspect of the present invention, the step of obtaining the damaged texture order parameter specifically includes: S101: Obtain the original image of the burn area, arrange the gradient sequence in the neighborhood of each pixel in order, analyze the change amplitude of adjacent gradient terms, filter the inflection point position and establish jump markers, and generate gradient jump distribution coefficients. S102: Based on the gradient jump distribution coefficient, compare the position of each pixel jump marker with the corresponding gradient sequence, analyze the jump rate distribution of each region, determine the degree of texture destruction and identify the destruction mode, and establish a region texture destruction index. S103: Based on the region texture destruction index, perform a unified mapping on the jump rate distribution, output texture order encoding, and output damaged texture order parameters.

[0007] As a further aspect of the present invention, the process of filtering inflection point positions and establishing transition markers specifically includes: After sorting the gradient sequences in the neighborhood of each pixel, the set of differences between adjacent gradient terms is obtained. The absolute value transformation of the set of differences between adjacent gradient terms is performed and sorted from small to large. The median difference of the set of differences between adjacent gradient terms is extracted as the difference center index, and the difference between the upper quantile difference and the lower quantile difference of the set of differences between adjacent gradient terms is used as the difference dispersion index. The difference center index and the difference dispersion index are added together to obtain the gradient jump judgment threshold. When the difference between adjacent gradient terms is greater than the gradient jump threshold and the difference between adjacent gradient terms shows a reversal of direction (one increasing and one decreasing or one decreasing and one increasing), the corresponding gradient sequence position is determined as an inflection point, and a jump marker is established at the target gradient sequence position.

[0008] As a further aspect of the present invention, the step of obtaining the degradation strength structural parameters specifically includes: S201: Based on the damaged texture order parameters, analyze the brightness change of each pixel in the same area, calculate the brightness change trend of each pixel with position, determine the change curve characteristics of each area, establish a brightness change trend sequence, and generate brightness change trend data. S202: Based on the brightness change trend data, perform difference calculation on the brightness change of each pixel and the heat loss related signal of the corresponding area, analyze the difference change trend in multiple directions, perform averaging processing on the difference change results in multiple directions, obtain the average rate of the multi-directional attenuation trend in the area, and generate the heat diffusion attenuation rate. S203: Based on the thermal diffusion attenuation rate and the average attenuation rate of the region, determine the intensity level of damage progression, establish a graded marker for the damage progression level, perform a structural transformation on the graded marker, and generate degradation intensity structural parameters.

[0009] As a further aspect of the present invention, the step of obtaining the bridging association parameters specifically includes: S301: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the candidate region of the wound edge, calculate the tension difference of each pair of adjacent pixels, screen the tension change locations and establish breakpoint markers to generate tension breakpoint distribution data. S302: Based on the tension breakpoint distribution data, classify the tension direction of the pixels on both sides of each breakpoint, determine the convergence characteristics of the tension direction, evaluate the continuous trend of the tension direction in space, and generate spatial trend analysis results. S303: Based on the spatial trend analysis results, the tension relationship is compared and deduced in the convergent region to form a bridging direction. Based on the tension stability of the pixels within the direction range, bridging connection points are selected, a cross-segment connection point sequence is constructed, and the bridging association parameters are generated.

[0010] As a further aspect of the present invention, the process of screening for tension abrupt change locations and establishing breakpoint markers specifically includes: Tension difference set is formed by tension data of adjacent pixels in the candidate region of the creation edge. The tension difference set is converted into absolute value and sorted in ascending order. The median difference of the sorted sequence is used as the tension difference benchmark, and the difference between the upper and lower quantile differences is used as the tension difference discrete value. The sum of the tension difference benchmark and the tension difference discrete value is set as the tension change judgment threshold. When the tension difference between adjacent pixels is greater than the tension change determination threshold and the corresponding tension change shows a reversal of direction (first increasing then decreasing or first decreasing then increasing) in adjacent positions, the position corresponding to the target adjacent pixel pair is determined as the tension change position, and a breakpoint mark is established at the target position.

[0011] As a further aspect of the present invention, the step of obtaining the burn location identification boundary specifically includes: S401: Based on the aforementioned bridging correlation parameters, analyze the color difference gradient density change of pixels in the corresponding region, calculate the enhancement and attenuation directions of consecutive pixels in the color difference density sequence, filter out points where the density direction is reversed at adjacent positions, establish pulse markers, and generate a color difference density pulse sequence. S402: Based on the color difference density pulse sequence, sort the distribution of pulse markers in the spatial sequence, analyze the directional changes of pulse points in the spatial distribution, determine the spatial order of the pulse distribution, and generate a boundary distribution sequence; S403: Based on the boundary distribution sequence, analyze the stable relationship of multiple markers in the direction change, select key points of the wound boundary, reorganize the continuous point set through the coordinates of the key points, and output them in spatial order to output the burn location identification boundary.

[0012] As a further aspect of the present invention, the step of obtaining the area scaling correction result specifically includes: S501: Based on the burn location identification boundary, analyze the curvature change of the corresponding skin area in the orthogonal direction, perform difference comparison on the curvature data in the orthogonal direction, determine the dominant relationship of curvature change in the area, and obtain the directional curvature difference parameter; S502: Based on the directional curvature difference parameter, determine the dominant direction and secondary direction of the region, perform stretching adjustment of the area projection on the dominant direction to compensate for the area shrinkage of the image, perform area shrinkage adjustment on the secondary direction to balance the expansion ratio of the region, and obtain the area direction adjustment parameter. S503: Based on the area direction adjustment parameters, the adjusted area mapping is organized into a structural output to generate area scaling correction results.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a texture order identifier is constructed by analyzing the pixel gradient sequence in the original image, a regional damage intensity judgment mechanism is established by combining brightness changes and thermal loss signal differences, the wound edge structure is deduced by the tension direction trend, the continuous boundary range is identified by combining color difference density, and the area unfolding comparison and correction are performed by fusing curvature differences. A systematic evaluation logic from boundary recognition to area mapping is established, the damage range is accurately delineated and matched with body surface features, and the accurate evaluation and regional unfolding of the burn area are achieved. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0015] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0017] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0018] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0019] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0021] Please see Figure 1 This invention provides an automatic burn area assessment method based on AI image recognition, comprising the following steps: S1: Call the original image of the burn area, analyze the gradient changes, sort the gradient sequence in the pixel neighborhood, compare the changes of adjacent gradient terms, filter inflection points and establish jump markers, combine them with the gradient sequence positions to form a jump rate distribution and perform mapping to form texture order encoding and generate damage texture order parameters. S2: Based on the damage texture order parameters, analyze the brightness changes in the corresponding area, evaluate the attenuation trend by calculating the difference between brightness and thermal loss signal, and take the average of the attenuation trend in multiple directions to determine the attenuation intensity and establish a degradation marker to generate degradation intensity structural parameters. S3: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the wound edge region, screen abrupt change points and establish breakpoint markers by comparing tension differences, perform convergence judgment on the tension directions on both sides of the breakpoint, deduce the bridging direction in the convergence region and screen stable points, and generate wound edge bridging association parameters. S4: Based on the bridging correlation parameters of the burn edge, analyze the color difference gradient density change in the corresponding area, identify the reversal point and establish pulse markers by comparing the enhancement and attenuation directions of the density sequence, sort the pulse markers and filter key points, construct a continuous boundary point set, and generate the burn location identification boundary. S5: Identify the boundary based on the burn location, analyze the curvature changes of the corresponding skin area in the orthogonal direction, determine the dominant and secondary directions by comparing the curvature differences, perform stretching adjustment on the area projection in the dominant direction, perform contraction adjustment on the secondary direction, organize the adjusted area mapping into structural output, and generate area stretching correction results.

[0022] Damage texture order parameters include gradient energy level jump distribution coefficient, texture order level encoding, and regional gradient sequence feature quantity; degradation intensity structure parameters include local brightness decay curve index, thermal damage progression level marker, and degradation structure partition index; wound edge bridging correlation parameters include tension breakpoint correlation coefficient, cross-segment bridging path index, and wound edge connectivity stability measure; burn location identification boundary includes color difference density pulse feature quantity, boundary key point coordinate set, and wound edge contour continuity index; area stretching correction results include orthogonal curvature direction ratio, principal and secondary direction area stretching factor, and body surface area mapping correction coefficient.

[0023] Please see Figure 2 The specific steps for obtaining the damaged texture order parameters are as follows: S101: Obtain the original image of the burn area, arrange the gradient sequence in the neighborhood of each pixel in order, analyze the change amplitude of adjacent gradient terms, filter the inflection point position and establish jump markers, and generate gradient jump distribution coefficients. After acquiring the original image of the burn area, the system first converts the original image of the burn area into a grayscale image to facilitate gradient calculation. For example, a burn image with a size of [missing information - likely a unit of measurement]... Pixels, one of which is a burned area is defined as A sub-region of a pixel; for this For each sub-region of a pixel, the system will process each pixel individually; that is, for any pixel within the burned area... Its neighborhood is defined as centered on that pixel. A pixel block of size 1.5, the neighborhood of which contains 8 neighboring pixels and a center pixel; the system calls the Sobel operator to calculate the horizontal gradient of each pixel in the neighborhood of this pixel. and vertical gradient And thus calculate the gradient magnitude of each pixel. For example, in a In the neighborhood of a pixel, the grayscale value of the center pixel is 150, the grayscale value of the pixel to its right is 180, and the grayscale value of the pixel below it is 160. Then the horizontal gradient... for Vertical gradient for gradient magnitude Approximately The system constructs a gradient sequence from these nine gradient magnitudes and arranges them in pixel order, for example, from left to right and from top to bottom. The process of obtaining the set of differences between adjacent gradient terms involves the system calculating the differences between all adjacent gradient terms in the gradient sequence. For example, if a gradient sequence is... Then the set of differences between adjacent gradient terms is ,Right now The system performs an absolute value transformation on the set of differences and sorts them in ascending order to obtain a sorted set of differences. The system extracts the median difference of the sorted set as the difference center index; for a set containing 8 elements, the median difference is the average of the 4th and 5th elements, i.e. The system uses the difference between the upper and lower quantiles of the set as the dispersion index; the upper quantile is the 75th percentile of the sorted set, i.e., the 6th element (8); the lower quantile is the 25th percentile of the sorted set, i.e., the 2nd element (5); the dispersion index is... The system adds the difference center index 6 and the difference dispersion index 3 to obtain the gradient jump judgment threshold. When the difference between adjacent gradient terms exceeds the gradient jump threshold of 9, and the difference between adjacent gradient terms shows a reversal of direction (one increasing and one decreasing or one decreasing and one increasing), the system determines the corresponding gradient sequence position as an inflection point and establishes a jump marker at the target gradient sequence position; for example, in the set of differences between adjacent gradient terms... In the process, if the difference is greater than the threshold of 9, and the differences before and after it are 4 and 7 respectively, with the direction changing from decreasing to increasing, the system determines the corresponding gradient sequence position as an inflection point and establishes a jump marker there. The system traverses the neighborhood of all pixels in the burned area, repeats the above gradient jump determination process, counts the number and position information of jump markers in the neighborhood of each pixel, and generates a gradient jump distribution coefficient, which reflects the activity level of the local texture. For each pixel in the burned area, the gradient jump distribution coefficient represents the frequency and position of significant gradient changes in its neighborhood.

[0024] S102: Based on the gradient jump distribution coefficient, compare the position of each pixel jump marker with the corresponding gradient sequence, analyze the jump rate distribution of each region, determine the degree of texture damage and identify the damage mode, and establish a region texture damage index. Process each pixel within the burned area; for example, for pixels within the burned area... Its gradient jump distribution coefficients contain the positions of jump markers in the neighborhood of that pixel; the system compares the positions of the jump markers in the neighborhood of that pixel with the positions of the corresponding gradient sequences, and calculates the jump rate of each pixel; the jump rate is calculated by dividing the number of jump markers in the neighborhood of that pixel by the total number of pixels in the neighborhood (usually 9); for example, if a If there are 3 transition markers in the neighborhood of a pixel, then the transition rate of that pixel is... The system aggregates these jump rate values ​​to construct a jump rate distribution for the burned area, representing the degree of texture disorder at different locations within the burned area. To determine the degree of texture damage, the system divides the jump rate distribution into three intervals: areas with a jump rate below 0.1 are classified as mild damage, areas with a jump rate between 0.1 and 0.4 are classified as moderate damage, and areas with a jump rate above 0.4 are classified as severe damage. For example, if the average jump rate of the burned area is 0.25, then the area belongs to moderate damage. When identifying damage patterns, the system judges based on the spatial clustering or dispersion of jump rates. If high jump rate areas exhibit a clustered distribution, it is identified as a "localized concentrated damage pattern"; if high... If the abrupt change rate regions exhibit a diffuse distribution, they are identified as a "uniform diffuse destruction mode." The system determines the clustering by calculating the size and number of connected components in high abrupt change rate regions: for example, if a connected component occupies more than 50% of the total area of ​​high abrupt change rate regions, it is determined as localized concentrated destruction; conversely, if multiple connected components are all less than 10% of the total area of ​​high abrupt change rate regions, it is determined as uniform diffuse destruction. The system integrates the degree of texture destruction and the identified destruction mode to establish a regional texture destruction index. This index is a composite value combining numerical values ​​and classification labels, such as "moderate destruction - localized concentration" or "severe destruction - uniform diffusion." This index provides a quantitative basis for subsequent texture order encoding.

[0025] S103: Based on the regional texture destruction index, perform a unified mapping on the jump rate distribution, output texture order encoding, and output damaged texture order parameters; A unified mapping is performed on the jump rate distribution of the burn area. This mapping process converts continuous jump rate values ​​into discrete texture order codes. The mapping rules are based on preset damage level standards, which are obtained through expert annotation and analysis of a large number of clinical burn images. For example, by analyzing the texture of images of 1,000 burn patients, the jump rate distribution characteristics corresponding to different burn depths are statistically analyzed, and cross-validation is performed to ensure the accuracy and stability of the coding. The specific mapping process is shown in Table 1, which defines the correspondence between jump rate intervals and texture order codes.

[0026] Table 1 Texture Order Encoding Mapping Table Jump rate range Texture order encoding Texture damage describe 001 minor damage The texture is clear, with a few local fluctuations. 010 Minor damage Textures begin to blur, and gradient variations increase in local areas. 011 Moderate damage The texture is significantly blurred, and large areas show frequent gradient changes. 100 Severe damage Texture structure is severely lost, and gradients change drastically in most areas. 101 Extremely severe damage Texture is completely lost, with irregular gradient distribution. As shown in Table 1, the system looks up the corresponding texture order code based on the average jump rate of the burned area. For example, if the average jump rate of a burned area is 0.32, according to Table 1, it will be mapped to texture order code 011, indicating moderate damage. If combined with the damage pattern identified in S102 (such as "moderate damage - local concentration"), the final texture order code will further include pattern information. For example, the system appends a pattern identifier after the 011 code, such as "011-L" (L represents local concentration) or "011-U" (U represents uniform dispersion). This damage texture order parameter, that is, the texture order code of the burned area, is a string, such as "011-L", which accurately quantifies the degree and pattern of texture damage in the burned area. This parameter serves as the input for subsequent steps, providing a structured representation of the texture features of the burned area. The advantage of this approach is that by mapping a continuous rate of change distribution to a discrete texture order encoding, the system can standardize and classify the texture features of the burned area, enabling texture damage of different burn degrees to be consistently identified and compared, thereby improving the stability and interpretability of the evaluation results.

[0027] Please see Figure 3 The specific steps for obtaining the degradation strength structural parameters are as follows: S201: Based on the damaged texture order parameters, analyze the brightness change of each pixel in the same area, calculate the brightness change trend of each pixel with position, determine the change curve characteristics of each area, establish a brightness change trend sequence, and generate brightness change trend data. Based on the damage texture order parameter, the corresponding burn area in the parameter is used as the analysis range; for example, if the burn area indicated by the damage texture order parameter is... The system analyzes the brightness value of each pixel within a sub-region of a pixel; the brightness change of each pixel is its grayscale value in the image; for example, a pixel's brightness value is 120 (range 0-255); to calculate the trend of brightness change of each pixel with position, the system constructs a sliding window, for example... The pixel size is adjusted and the pixel slides within the burned area in 1-pixel increments. For each center pixel in the window, the system calculates a weighted average of its brightness value and the brightness values ​​of its surrounding pixels, and obtains its trend by comparing it with the brightness value of the center pixel itself. Specifically, the trend is calculated by fitting a quadratic polynomial to the local brightness change; for example, in a... In the window, the brightness value can be represented as System Fitting ,in These are the fitting coefficients; through The system uses the sign and size of the value to determine whether the local trend of brightness change is convex, concave, or flat; for example, if... All values ​​are positive and relatively large, indicating that the brightness is increasing in this direction, representing a region with high brightness; if All values ​​are negative and relatively large, indicating a decreasing trend in brightness in this direction, representing a region with low brightness; if A value close to 0 indicates that the brightness shows a gradual trend in this direction, representing uniform brightness in the area; the system combines the brightness variation trend of each pixel (e.g., convexity, concavity, flatness) and its fitting coefficient. The system stores information such as brightness change trend sequence to form a brightness change trend sequence. To determine the change curve characteristics of each region, the system performs statistical analysis on the brightness change trend sequence of all pixels within the burn area. For example, it counts the proportion of raised, recessed, and flat pixels, as well as their spatial distribution. If more than 70% of the pixels in the burn area show a recessed trend, the system determines the change curve characteristic of the area as "overall recessed"; if more than 70% of the pixels show a raised trend, the system determines it as "overall raised"; if all trends are evenly distributed, it is determined as "mixed characteristics". Finally, the system generates brightness change trend data, which includes the brightness change trend of each pixel, the fitting coefficient, and the overall change curve characteristics of the region. This data provides basic brightness characteristics for subsequent assessment of damage progression intensity.

[0028] S202: Based on the brightness change trend data, perform difference calculation on the brightness change of each pixel and the heat loss related signal of the corresponding area, analyze the difference change trend in multiple directions, perform averaging processing on the difference change results in multiple directions, obtain the average rate of the multi-directional attenuation trend in the area, and generate the heat diffusion attenuation rate. Each pixel within the burned area is processed; the brightness change of each pixel is determined by the fitting coefficient in S201. Once determined, it represents the local brightness characteristics; simultaneously, the system acquires the thermal loss correlation signal corresponding to that pixel; the thermal loss correlation signal is usually acquired by an infrared thermal imager, which reflects the temperature distribution on the tissue surface. For example, in a burn area, reduced tissue activity leads to local temperature increases or decreases, and these temperature anomalies can be detected by an infrared thermal imager; for example, assuming the brightness fitting coefficient of a pixel is... (Indicating a slight localized brightness spike), the corresponding temperature value measured by the infrared thermal imager is 38.5 degrees Celsius; the system performs a difference calculation between the brightness fitting coefficient and the thermal loss signal, for example, the difference calculation is defined as: ,in It is the brightness-to-temperature conversion coefficient. This is a reference body temperature (usually 37 degrees Celsius); The value is determined based on clinical trial data. For example, by performing infrared thermography and brightness analysis on burn tissues of different depths, a mapping relationship between brightness and temperature is established, and regression analysis is performed. For instance, if the average change in the brightness fitting coefficient is 0.01 when the temperature increases by 1 degree Celsius, then... In this example, the result of the difference operation is: The system analyzes the variation trends of differences in multiple directions by calculating the difference gradient of each pixel in eight directions, including horizontal, vertical, and diagonal. For example, for a pixel... The gradient of its horizontal differences is The vertical difference gradient is The average processing is performed on the results of the multi-directional difference changes, that is, the average of the difference gradient magnitudes in all directions is taken to obtain the average rate of the multi-directional decay trend within the region; for example, for a pixel, the difference gradient magnitudes in its eight directions are as follows: The average rate is the average of these values, approximately The system statistically analyzes the average rate of all pixels within the burned area to generate a thermal diffusion attenuation rate. This rate is a numerical value that reflects the average speed of heat diffusion within the burned area and the combined attenuation trend of brightness and heat loss.

[0029] S203: Based on the thermal diffusion decay rate and the average decay rate of the region, determine the intensity level of damage progression, establish a graded label for the damage progression level, perform a structural transformation on the graded label, and generate degradation intensity structural parameters. The intensity level of damage progression is determined based on the thermal diffusion attenuation rate and the average attenuation rate of the burn area. The average attenuation rate of the area is obtained by averaging the thermal diffusion attenuation rates of all pixels within the burn area. For example, if the thermal diffusion attenuation rates of pixels within the burn area are distributed within a certain range, the system calculates their average value as the average attenuation rate of the area. To determine the intensity level of damage progression, the system presets a set of grading thresholds for damage progression levels. These thresholds are set based on the correlation between clinical burn depth and thermal diffusion attenuation rate. For example, by measuring the thermal diffusion attenuation rate of burn tissues with different burn depths (first degree, superficial second degree, deep second degree, and third degree) and having experts conduct depth assessments, a correspondence between attenuation rate and depth is established to ensure the clinical effectiveness of the grading thresholds. As shown in Table 2, this table defines the correspondence between thermal diffusion attenuation rate and damage progression level.

[0030] Table 2 Thresholds for Classification of Damage Progression Level Average decay rate range Damage progression level describe (−∞,−0.01] Level 0 Tissue recovery or no signs of damage (−0.01,0.02] Level 1 The injury progressed from mild to severe, primarily involving epidermal damage (first-degree burns). (0.02,0.05] Level 2 Moderate injury progressed, involving part of the dermis (superficial second-degree burn). (0.05,0.08] Level 3 The injury progressed to a severe stage, involving deep dermis (deep second-degree burns). Level 4 The injury progressed to extremely severe levels, involving the entire thickness of the skin (third-degree burns). As shown in Table 2, the system compares the average decay rate of the burn area with a preset threshold to determine the intensity level of damage progression and establish a grading label for the damage progression level. For example, if the average decay rate of the burn area is 0.065, according to Table 2, this rate falls into... (0.05,0.08] The system establishes a grading label of "Grade 3" to indicate severe injury progression. The system performs a structured transformation on the grading label to generate a degradation intensity structural parameter. This transformation typically converts the grading label into a numerical value or a string with a specific format; for example, "Grade 3" can be converted to the numerical value 3 or the string "DI_3" (DamageIntensity_3). This degradation intensity structural parameter, a numerical value or string such as "DI_3," quantifies the intensity of burn damage progression. This parameter serves as input for subsequent steps, providing quantitative information on burn depth and severity. The advantage of this approach is that by mapping complex thermal diffusion decay rates to clear injury progression grades, the system can provide intuitive and standardized burn depth assessments, improving the efficiency and accuracy of clinical assessments and providing a clear basis for treatment selection.

[0031] Please see Figure 4 The specific steps for obtaining the connection parameters of the Chuangyuan bridging are as follows: S301: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the candidate region of the wound edge, calculate the tension difference of each pair of adjacent pixels, screen the tension change locations and establish breakpoint markers to generate tension breakpoint distribution data. Based on the degradation intensity structural parameters, the severity of burn damage is identified, and candidate wound edge regions are determined accordingly. For example, if the degradation intensity structural parameter is "DI_3", indicating a deep second-degree burn, the system will use a pre-set clinical experience knowledge base to select areas where wound edges may exist as candidate wound edge regions. These areas are typically the boundary between the burn area and the normal skin area, such as a ring-shaped area with a width of 10-20 pixels. Within the candidate wound edge regions, the system analyzes the tension changes of adjacent pixels. Pixel tension data is obtained through image morphological analysis, such as by calculating local texture directionality and anisotropy, or by simulating the skin's stress state using a finite element model. For example, the tension value of a pixel can be represented as the average force exerted on the skin at that point, in Newtons per millimeter (N / mm²), and in practical applications, it is quantized to a grayscale value range of 0-255. The system calculates the tension difference between each pair of adjacent pixels. For example, for pixels in the candidate wound edge regions... The system calculates the tension difference between the pixel and its eight neighboring pixels; if the pixel The tension value Its right pixel The tension value The tension difference is The system constructs a tension difference set from these tension difference values. To filter locations of sudden tension changes and establish breakpoint markers, the system performs an absolute value transformation on the tension difference set and sorts it in ascending order. For example, the tension difference set is... (Unit: grayscale value); The system uses the median difference of the sorted sequence as the benchmark for tension difference; For a set of 8 elements, the median difference is... The difference between the upper and lower quantiles of the system is used as the tension difference dispersion; the upper quantile is the 75th percentile, i.e., the 6th element 25; the lower quantile is the 25th percentile, i.e., the 2nd element 8; the tension difference dispersion is... The system sets the sum of the tension difference baseline of 17.5 and the tension difference discrete value of 17 as the threshold for determining tension abrupt changes. When the tension difference between adjacent pixels exceeds the tension abrupt change threshold of 34.5, and the corresponding tension change shows a reversal of direction (increasing first and then decreasing or decreasing first and then increasing) in adjacent positions, the system determines the position corresponding to the target adjacent pixel pair as the tension abrupt change position and establishes a breakpoint marker at the target position. For example, if the tension difference between an adjacent pixel pair is 35, and the tension differences before and after it are 30 and 28 respectively (showing an increase followed by a decrease), then this position is determined as the tension abrupt change position and a breakpoint marker is established. The system traverses the candidate edge region, repeats the above screening process, and generates tension breakpoint distribution data, which contains the coordinates of all tension abrupt change positions and breakpoint markers.

[0032] S302: Based on the tension breakpoint distribution data, classify the tension direction of the pixels on both sides of each breakpoint, determine the convergence characteristics of the tension direction, evaluate the continuous trend of the tension direction in space, and generate spatial trend analysis results. Based on the tension breakpoint distribution data, each breakpoint location is processed; for example, a breakpoint marker is located at a pixel. The system needs to obtain the tension direction of the pixels on both sides of the breakpoint. The tension direction can be obtained by calculating the local gradient direction within the pixel's neighborhood or by principal component analysis (PCA). For example, for 10 pixels on each side of the breakpoint, the system calculates their tension direction, representing it as an angle value from 0 to 360 degrees. The system then categorizes the tension directions of the pixels on both sides of each breakpoint, grouping pixels with similar directions together. For example, the system sets a direction deviation threshold, such as 15 degrees; if the tension directions of two pixels differ by less than 15 degrees, they are considered to belong to the same direction category. For example, the average tension direction of the pixels to the left of the breakpoint is... The average tension direction of the pixel to the right of the breakpoint is The system determines the convergence of tension directions, i.e., whether the tension directions on both sides of the break point belong to the same category; if the average direction difference on both sides is... If the average direction difference is less than 15 degrees, it is considered "convergent"; conversely, if the average direction difference is greater than the threshold, it is considered "non-convergent". For example, if the tension directions on both sides of a break point are convergent, it indicates that the break point may not be a true wound fracture, but rather a stress concentration within the tissue. The system evaluates the continuous trend of tension direction in space by analyzing the convergence of tension directions between multiple adjacent break points. For example, if the system identifies 5 adjacent break points in a region, 3 of which show convergent tension directions on both sides, while the other 2 do not, the system considers convergent break points as potential path points, and non-convergent break points as true tissue fractures. The system constructs a graph... The system uses a structure to represent the connection relationship between breakpoints, where nodes are breakpoints and edges represent the convergence of tension directions. It uses Dijkstra's algorithm to find convergent paths and evaluate the continuous trend of tension directions in space. For example, if a continuous path consisting of multiple converging breakpoints exists at the boundary of the burn area, its tension direction shows a smooth transition, and the system evaluates it as having a "strong continuous trend." If the tension directions differ greatly between breakpoints, it is evaluated as having a "weak continuous trend." Finally, the system generates a spatial trend analysis result, which includes the classification of tension directions on both sides of each breakpoint, the convergence judgment, and the overall continuous trend assessment of tension directions within the region, providing spatial context for subsequent bridging directions.

[0033] S303: Based on the spatial trend analysis results, the tension relationship is compared and deduced in the convergent region to form a bridging direction. Based on the tension stability of the pixels within the direction range, the bridging connection points are selected, a cross-segment connection point sequence is constructed, and the bridging association parameters are generated. Based on spatial trend analysis, convergent regions within the candidate wound edge area are identified. These regions exhibit convergent tension directions on both sides of the breakpoints. For example, if spatial trend analysis shows that multiple adjacent breakpoints within a continuous pixel region exhibit convergence, the system defines this region as a convergent region. Within this convergent region, the system compares and derives tension relationships to form bridging directions. This comparison and derivation process involves analyzing the tension gradient direction and magnitude of pixels within the convergent region to find smooth transition areas of tension changes. For example, within the convergent region, the system calculates the tension gradient value every 5 pixels along a path where tension directions converge, and compares the differences between these gradient values. If the difference is within a set threshold (e.g., the tension gradient change is less than 20% of the average gradient change), a bridging direction is derived, meaning these points are considered to be "bridged," forming a potential wound edge connection line. For example, the starting point of a bridging direction is a pixel... The endpoint is a pixel. The system filters bridging connection points based on the tension stability of pixels within the pointing range; tension stability is evaluated by calculating the standard deviation of tension values ​​within the pixel's neighborhood. In the neighborhood of a pixel, if the standard deviation of the tension value is less than a set stability threshold (e.g., the standard deviation is less than 10 grayscale values), the pixel is considered to have stable tension. The system prioritizes pixels with stable tension that are located on the bridging direction as connection points. For example, on a bridging direction line, the system selects a stable tension point every 3 pixels as a candidate connection point. The system constructs these selected connection points into a cross-segment connection point sequence, for example, the sequence contains connection points. Finally, the system generates wound edge bridging correlation parameters, which contain the coordinate information of all cross-segment connection point sequences, such as a list or array where each element represents the coordinates of a connection point and is arranged in spatial order. This parameter provides a preliminary connection framework for the subsequent accurate construction of wound edge boundaries. The advantage of this approach is that by combining the convergence of tension directions and tension stability, the system can effectively identify potential connection paths at burn wound edges, avoiding misjudging tissue fractures as continuous areas, thereby improving the accuracy and reliability of wound edge identification.

[0034] Please see Figure 5 The specific steps for obtaining the boundary of burn location identification are as follows: S401: Based on the bridging correlation parameters, analyze the color difference gradient density change of pixels in the corresponding region, calculate the enhancement and attenuation directions of continuous pixels in the color difference density sequence, filter the points where the density direction is reversed in adjacent positions, establish pulse markers, and generate a color difference density pulse sequence. Based on the bridging correlation parameters, a preliminary sequence of bridging points is identified, defining the corresponding regions for subsequent analysis. For example, if the bridging correlation parameters contain a series of bridging points, the system will use these bridging points as the center and extend outwards to a certain range (e.g., 5-10 pixels) as the analysis range. Within this region, the system analyzes the color difference gradient density change of each pixel. Color difference gradient density is an indicator that measures the intensity and density of color changes. It is calculated by converting the image from the RGB color space to the Lab color space and then calculating the color difference gradient density of each pixel. The gradient magnitudes of the components are calculated and then weighted to obtain the sum; for example, the Lab value of a pixel is... When the Lab value of the surrounding pixels changes, a color difference gradient is generated; the color difference gradient density is calculated as follows: And these gradient values ​​are statistically analyzed within a local window; for example, in a Within a window, if the color difference gradient values ​​between all pixels are very large, then the color difference gradient density of that window is high. The system calculates the enhancement and decay directions of consecutive pixels in the color difference density sequence. For example, the system extracts the color difference gradient density values ​​of pixels along a preset scanning path (such as horizontal or vertical scanning) to form a sequence. The system compares the differences between adjacent density values ​​in the sequence to determine whether they are increasing (enhancement direction) or decreasing (decrease direction). For example, if a sequence is... Then its direction change sequence is The system filters points where the density reverses direction at adjacent locations and establishes pulse markers; direction reversal refers to a change from enhancement to attenuation, or from attenuation to enhancement; for example, in a direction change sequence, from arrive The position is the direction reversal point, from arrive The location is also the direction reversal point; the system uses these direction reversal points as pulse markers because they usually indicate sudden changes in texture or color boundaries; the system records all identified pulse markers, including their coordinates and corresponding color difference gradient density values, and generates a color difference density pulse sequence; this sequence is the original feature point set for boundary identification.

[0035] S402: Based on the color difference density pulse sequence, sort the distribution of pulse markers in the spatial sequence, analyze the directional changes of pulse points in the spatial distribution, determine the spatial order of the pulse distribution, and generate a boundary distribution sequence; Based on the color difference density pulse sequence, the distribution of pulse markers in the spatial sequence is sorted. For example, if the pulse sequence contains a large amount of pulse marker coordinate information, the system first sorts the pixels in ascending order based on their X coordinates. If the X coordinates are the same, it then sorts them in ascending order based on their Y coordinates, thus obtaining a spatially ordered list of pulse markers. The system analyzes the directional changes of pulse points in the spatial distribution by constructing local neighborhoods (e.g., a...). The system uses a pixel-based window to observe the arrangement trend of the pulse markers; within each local neighborhood, the system calculates the average direction vector of the pulse markers; for example, if the pulse markers are mainly arranged horizontally, their direction vector is close to... or If arranged vertically, the direction vectors are close to... or The system determines the spatial order of pulse distribution by analyzing the continuity and consistency of these local direction vectors. For example, the system sets a direction consistency threshold (e.g., the average direction vector deviation of five consecutive local neighborhoods is less than a certain value). If, within a certain region, the pulse direction vectors of multiple local neighborhoods exhibit high consistency, the pulse distribution in that region is determined to be "high-order," indicating the presence of a clear boundary structure. If the direction vectors are disordered, it is determined to be "low-order," indicating internal chaos within the region. The system quantifies the degree of order by calculating the average cosine value of the local direction vectors. For example, if the average cosine value is close to 1, the order is high; if it is close to 0, the order is low. The system combines these order judgment results with the coordinate information of the pulse markers to generate a boundary distribution sequence. This sequence contains the spatial location of the pulse markers, local direction vectors, and regional order evaluation results, providing spatially structured information for the subsequent screening of key points of the boundary.

[0036] S403: Based on the boundary distribution sequence, analyze the stable relationship of multiple markers in the direction change, screen out the key points of the wound boundary, reorganize the continuous point set through the coordinates of the key points, and output them in spatial order to output the burn location identification boundary. Based on the boundary distribution sequence, the stable relationship of multiple pulse markers in the directional change of the pulse markers is analyzed. A stable relationship refers to the consistent direction vector and high order exhibited by multiple consecutive pulse markers in the boundary distribution sequence. For example, if the boundary distribution sequence shows that the local direction vector deviation of 15 consecutive pulse markers within a certain region is less than [a certain value], then [the analysis is conducted]. If the spatial order of these markers is assessed as "high order," the system determines that a stable relationship exists between them. The system quantifies this stable relationship by setting a stability threshold (e.g., containing at least 10 consecutive pulse markers and having a local orientation consistency higher than 90%). The system selects key points at the boundary of the stable relationship; these key points are the start and end points of the stable relationship sequence, as well as points with relatively large orientation changes. For example, in a stable pulse sequence, the system selects a point as a key point at regular intervals (e.g., every 5 pulse markers), and selects the start and end points of the sequence as key points. Furthermore, if the local orientation vector of a point in the sequence differs from the orientation vectors of the points before and after it by more than [a certain percentage], the system determines that a stable relationship exists between these markers. Even if the region is generally stable, this point will be marked as a key point because it may indicate a boundary inflection. The system reconstructs a continuous set of points using the key point coordinates. The reconstruction process employs spline interpolation or Bézier curve fitting; for example, using cubic spline interpolation, a smooth and continuous curve is generated from the selected key points. This curve connects all key points, forming a continuous boundary of the wound edge. The system outputs the reconstructed continuous set of points in spatial order, i.e., arranging the coordinates of all boundary points from the start point to the end point. Finally, the system outputs the burn location identification boundary, which is an ordered list containing the coordinates of all boundary points of the wound edge, for example, a two-dimensional array. ,in This parameter represents the number of boundary points; it provides accurate wound edge information for subsequent area calculation and correction. The advantage of this approach is that by conducting in-depth analysis of the spatial distribution and directional changes of pulse markers, combining stable relationships to screen key points and perform smooth interpolation, the system can generate highly accurate and continuous wound edge boundaries, significantly improving the reliability and accuracy of burn area assessment.

[0037] Please see Figure 6 The specific steps for obtaining the area scaling correction results are as follows: S501: Based on the boundary identification of burn location, analyze the curvature changes of the corresponding skin area in the orthogonal direction, perform difference comparison on the curvature data in the orthogonal direction, determine the dominant relationship of curvature changes in the area, and obtain the directional curvature difference parameters; Based on the burn location identification boundary, the precise wound edge boundary line is identified. For example, the burn location identification boundary is a list containing multiple ordered coordinate points, which constitute the geometric boundary of the burn area. The system uses this boundary line as a basis to determine the corresponding skin region within it as the analysis object. Within this skin region, the system analyzes its curvature changes in orthogonal directions. Curvature is a geometric quantity that measures the degree of bending of an object, usually obtained through 3D image reconstruction techniques, such as generating 3D point cloud data of the skin surface through structured light scanning or multi-view stereo matching, then calculating the surface normal vector at each point, and further calculating the principal curvature. and The curvature in the orthogonal direction is the principal curvature. and For example, in a burn area, after a pixel is reconstructed on the skin surface, its two principal curvatures are respectively... and The system performs a difference comparison on the curvature data in orthogonal directions, i.e., a comparison. The value of the difference indicates that the region exhibits a greater difference in curvature between the two orthogonal directions, and a more significant geometric deformation. For example, if... The system determines the dominant relationship of curvature changes within a region by setting a curvature difference threshold (e.g., If the difference value is greater than this threshold, then a dominant bending direction is considered to exist; for example, if the difference value... Greater than the threshold If so, the system determines that there is a dominant curvature direction, where The corresponding direction is the dominant direction because its absolute value is larger; the system will consider these comparison results, including The values ​​of , and the differences between them, together with the determination of the dominant relationship, together yield the directional curvature difference parameter; this parameter is a dataset containing curvature information of each pixel and the determination result of the dominant direction, providing a basis for three-dimensional geometric deformation for subsequent area correction.

[0038] S502: Based on the directional curvature difference parameter, determine the dominant and secondary directions of the region, perform area projection stretching adjustment on the dominant direction to compensate for the area shrinkage of the image, perform area shrinkage adjustment on the secondary direction, balance the expansion ratio of the region, and obtain the area direction adjustment parameter. Based on the directional curvature difference parameter, for each pixel in the burned skin area, its dominant and secondary directions are determined; for example, if the curvature difference parameter of a pixel shows that the dominant curvature is... The corresponding direction is the dominant direction, and the curvature value is... The principal curvature The corresponding direction is the secondary direction, and the curvature value is ; Perform area projection stretching adjustment on the dominant direction to compensate for image area shrinkage; compensation coefficient The calculation is related to the dominant curvature, for example, the compensation coefficient. ,in It involves adjusting the sensitivity coefficient (determined through clinical trials, for example...) ), It is the absolute value of curvature in the dominant direction. It is the average depth of the burned area (obtained via S203, for example) For example, in a burn area, the average dominant curvature is... Then the compensation coefficient This means that in the dominant direction, the projected area of ​​each pixel needs to be stretched by 1.5%; area contraction adjustments are performed in the secondary directions to balance the expansion ratio of the region; contraction coefficient. The calculation is related to the secondary curvature, for example, the contraction coefficient. ,in It is to adjust the sensitivity coefficient (e.g.) ), It is the absolute value of curvature in the secondary direction; for example, the average curvature in the secondary direction is Then the shrinkage coefficient This means that in the secondary direction, the projected area of ​​each pixel needs to shrink by 0.2%; when calculating the area, the system will adjust the two-dimensional projected area of ​​each pixel according to its local... and The coefficients are adjusted; for example, if the initial projected area of ​​a pixel is... After adjustment, its true area can be estimated as The system integrates these adjustment coefficients and the adjusted local area information to obtain the area direction adjustment parameter; this parameter provides local area information corrected for three-dimensional deformation for the accurate calculation of the final burn area.

[0039] S503: Based on the area orientation adjustment parameters, the adjusted area mapping is organized into structural output to generate area scaling correction results; Based on the area orientation adjustment parameters, the system includes local area information for each pixel within the burn area after stretching adjustment in the dominant direction and contraction adjustment in the secondary direction. For example, the area orientation adjustment parameters are a two-dimensional array, where each element represents a pixel and stores the actual local area value of that pixel after adjustment. The system accumulates these adjusted local area values ​​to calculate the total area of ​​the entire burn area. For example, if the burn area contains 10,000 pixels, the adjusted area of ​​each pixel is as follows: The total area of ​​the burn zone Assume the average area of ​​each pixel after adjustment is The total area is The system maps and organizes this final burn area value into a structural output, typically a floating-point number representing the burn area in millimeters squared or as a percentage. The structural output may also include metadata such as area units and correction factors. For example, the structural output might be... Ultimately, the system generates an area scaling correction result. This parameter is a structured data object that accurately represents the burn area after three-dimensional curvature deformation correction. This parameter can be directly used for clinical assessment, providing doctors with accurate burn area data. The advantage of this approach is that by performing refined three-dimensional curvature deformation correction on the image projection area, the system can overcome the limitations of traditional two-dimensional image assessment, making the calculated burn area closer to the true surface area of ​​the patient's skin. This improves the accuracy and reliability of burn area assessment, providing more precise data support for clinical diagnosis and treatment decisions.

[0040] 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. An automatic burn area assessment method based on AI image recognition, characterized in that, Includes the following steps: S1: Call the original image of the burn area, analyze the gradient changes, sort the gradient sequence in the pixel neighborhood, compare the changes of adjacent gradient terms, filter inflection points and establish jump markers, combine them with the gradient sequence positions to form a jump rate distribution and perform mapping to form texture order encoding and generate damage texture order parameters. S2: Based on the damage texture order parameters, analyze the brightness changes in the corresponding area, evaluate the attenuation trend by calculating the difference between brightness and thermal loss signal, and take the average of the attenuation trend in multiple directions to determine the attenuation intensity and establish a degradation marker to generate degradation intensity structural parameters. S3: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the wound edge region, screen abrupt change points and establish breakpoint markers by comparing tension differences, perform convergence judgment on the tension directions on both sides of the breakpoint, deduce the bridging direction in the convergence region and screen stable points, and generate wound edge bridging association parameters. S4: Based on the aforementioned wound bridging correlation parameters, analyze the color difference gradient density change in the corresponding region, identify the reversal point and establish pulse markers by comparing the enhancement and attenuation directions of the density sequence, sort the pulse markers and filter key points, construct a continuous boundary point set, and generate the burn location identification boundary.

2. The automatic burn area assessment method based on AI image recognition according to claim 1, characterized in that, The damage texture order parameters include gradient energy level jump distribution coefficient, texture order level encoding, and regional gradient sequence feature quantity. The degradation intensity structure parameters include local brightness decay curve index, thermal damage progression level marker, and degradation structure partition index. The wound edge bridging association parameters include tension breakpoint association coefficient, cross-segment bridging path index, and wound edge connectivity stability measure. The burn location identification boundary includes color difference density pulse feature quantity, boundary key point coordinate set, and wound edge contour continuity index.

3. The automatic burn area assessment method based on AI image recognition according to claim 1, characterized in that, The specific steps for obtaining the damaged texture order parameters are as follows: S101: Obtain the original image of the burn area, arrange the gradient sequence in the neighborhood of each pixel in order, analyze the change amplitude of adjacent gradient terms, filter the inflection point position and establish jump markers, and generate gradient jump distribution coefficients. S102: Based on the gradient jump distribution coefficient, compare the position of each pixel jump marker with the corresponding gradient sequence, analyze the jump rate distribution of each region, determine the degree of texture destruction and identify the destruction mode, and establish a region texture destruction index. S103: Based on the region texture destruction index, perform a unified mapping on the jump rate distribution, output texture order encoding, and output damaged texture order parameters.

4. The automatic burn area assessment method based on AI image recognition according to claim 3, characterized in that, The process of filtering inflection point locations and establishing transition markers is as follows: After sorting the gradient sequences in the neighborhood of each pixel, the set of differences between adjacent gradient terms is obtained. The absolute value transformation of the set of differences between adjacent gradient terms is performed and sorted from small to large. The median difference of the set of differences between adjacent gradient terms is extracted as the difference center index, and the difference between the upper quantile difference and the lower quantile difference of the set of differences between adjacent gradient terms is used as the difference dispersion index. The difference center index and the difference dispersion index are added together to obtain the gradient jump judgment threshold. When the difference between adjacent gradient terms is greater than the gradient jump threshold and the difference between adjacent gradient terms shows a reversal of direction (one increasing and one decreasing or one decreasing and one increasing), the corresponding gradient sequence position is determined as an inflection point, and a jump marker is established at the target gradient sequence position.

5. The automatic burn area assessment method based on AI image recognition according to claim 3, characterized in that, The specific steps for obtaining the degradation strength structural parameters are as follows: S201: Based on the damaged texture order parameters, analyze the brightness change of each pixel in the same area, calculate the brightness change trend of each pixel with position, determine the change curve characteristics of each area, establish a brightness change trend sequence, and generate brightness change trend data. S202: Based on the brightness change trend data, perform difference calculation on the brightness change of each pixel and the heat loss related signal of the corresponding area, analyze the difference change trend in multiple directions, perform averaging processing on the difference change results in multiple directions, obtain the average rate of the multi-directional attenuation trend in the area, and generate the heat diffusion attenuation rate. S203: Based on the thermal diffusion attenuation rate and the average attenuation rate of the region, determine the intensity level of damage progression, establish a graded marker for the damage progression level, perform a structural transformation on the graded marker, and generate degradation intensity structural parameters.

6. The automatic burn area assessment method based on AI image recognition according to claim 5, characterized in that, The specific steps for obtaining the bridging association parameters are as follows: S301: Based on the degradation intensity structural parameters, analyze the tension changes of adjacent pixels in the candidate region of the wound edge, calculate the tension difference of each pair of adjacent pixels, screen the tension change locations and establish breakpoint markers to generate tension breakpoint distribution data. S302: Based on the tension breakpoint distribution data, classify the tension direction of the pixels on both sides of each breakpoint, determine the convergence characteristics of the tension direction, evaluate the continuous trend of the tension direction in space, and generate spatial trend analysis results. S303: Based on the spatial trend analysis results, the tension relationship is compared and deduced in the convergent region to form a bridging direction. Based on the tension stability of the pixels within the direction range, bridging connection points are selected, a cross-segment connection point sequence is constructed, and the bridging association parameters are generated.

7. The automatic burn area assessment method based on AI image recognition according to claim 6, characterized in that, The process of screening for locations of tension abrupt changes and establishing breakpoint markers is specifically as follows: Tension difference set is formed by tension data of adjacent pixels in the candidate region of the creation edge. The tension difference set is converted into absolute value and sorted in ascending order. The median difference of the sorted sequence is used as the tension difference benchmark, and the difference between the upper and lower quantile differences is used as the tension difference discrete value. The sum of the tension difference benchmark and the tension difference discrete value is set as the tension change judgment threshold. When the tension difference between adjacent pixels is greater than the tension change determination threshold and the corresponding tension change shows a reversal of direction (first increasing then decreasing or first decreasing then increasing) in adjacent positions, the position corresponding to the target adjacent pixel pair is determined as the tension change position, and a breakpoint mark is established at the target position.

8. The automatic burn area assessment method based on AI image recognition according to claim 6, characterized in that, The specific steps for obtaining the burn location identification boundary are as follows: S401: Based on the aforementioned bridging correlation parameters, analyze the color difference gradient density change of pixels in the corresponding region, calculate the enhancement and attenuation directions of consecutive pixels in the color difference density sequence, filter out points where the density direction is reversed at adjacent positions, establish pulse markers, and generate a color difference density pulse sequence. S402: Based on the color difference density pulse sequence, sort the distribution of pulse markers in the spatial sequence, analyze the directional changes of pulse points in the spatial distribution, determine the spatial order of the pulse distribution, and generate a boundary distribution sequence; S403: Based on the boundary distribution sequence, analyze the stable relationship of multiple markers in the direction change, select key points of the wound boundary, reorganize the continuous point set through the coordinates of the key points, and output them in spatial order to output the burn location identification boundary.

9. The automatic burn area assessment method based on AI image recognition according to claim 1, characterized in that, The method further includes: S5: Identify the boundary based on the burn location, analyze the curvature change of the corresponding skin area in the orthogonal direction, determine the dominant and secondary directions by comparing the curvature differences, perform stretching adjustment on the area projection in the dominant direction, perform contraction adjustment on the secondary direction, organize the adjusted area mapping into structural output, and generate area stretching correction results. The area stretching correction results include the ratio of orthogonal curvature directions, the area stretching factor in the primary and secondary directions, and the surface area mapping correction coefficient.

10. The automatic burn area assessment method based on AI image recognition according to claim 9, characterized in that, The specific steps for obtaining the area scaling correction result are as follows: S501: Based on the burn location identification boundary, analyze the curvature change of the corresponding skin area in the orthogonal direction, perform difference comparison on the curvature data in the orthogonal direction, determine the dominant relationship of curvature change in the area, and obtain the directional curvature difference parameter; S502: Based on the directional curvature difference parameter, determine the dominant direction and secondary direction of the region, perform stretching adjustment of the area projection on the dominant direction to compensate for the area shrinkage of the image, perform area shrinkage adjustment on the secondary direction to balance the expansion ratio of the region, and obtain the area direction adjustment parameter. S503: Based on the area direction adjustment parameters, the adjusted area mapping is organized into a structural output to generate area scaling correction results.

Images