Method and system for real-time monitoring of burn wound healing process

Abstract

The present application relates to the technical field of intelligent medical monitoring, in particular to a burn wound healing process real-time monitoring method and system, comprising the following steps: collecting oxygen saturation data to construct a microcirculation blood flow feature set, performing texture tracking on a wound image to construct a global deformation vector field, fitting an edge line and combining the deformation field to obtain a net proliferation migration vector field, constructing a recursive feature matrix and calculating a laminar flow property and a capture time index, and outputting an evaluation result, in the present application, by combining microcirculation blood flow features and deformation field mapping, background motion interference is eliminated, net proliferation migration behavior is accurately quantified, a recursive feature matrix is constructed by mapping time series signals to phase space, laminar flow property and capture time index in state evolution are identified, multi-dimensional state determination is performed by fusing physiological blood flow parameters and morphological contraction coefficients, non-contact objective quantitative analysis of the healing process is realized, multi-dimensional pathological grading standards are established, and the dynamic trend of tissue remodeling is accurately reflected.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical monitoring technology, and in particular to a method and system for real-time monitoring of the healing process of burn wounds. Background Technology

[0002] The field of intelligent medical monitoring technology mainly encompasses a comprehensive technical system that utilizes the Internet of Things, biosensors, computer vision, and data analysis technologies to sense, transmit, and process human physiological and pathological indicators. The core aspects of this field involve collecting vital sign data through contact or non-contact devices, aggregating information via communication networks, and assessing health status using a computing platform. The aim is to build a digital system that can assist in clinical diagnosis and treatment management. Traditional real-time monitoring methods for burn wound healing refer to the technology of tracking the dynamic recovery of skin tissue after a burn, including stages such as inflammatory response, cell proliferation, and tissue remodeling. This typically involves medical personnel periodically observing the wound color, distribution of necrotic tissue, and exudate characteristics with the naked eye; using palpation on the back of the hand to determine skin temperature and the degree of redness and swelling around the wound; collecting secretion samples with sterile cotton swabs for microbial culture; and using surgical instruments to cut small amounts of granulation tissue for pathological section examination to obtain biological evidence reflecting the degree of wound healing.

[0003] Traditional methods for real-time monitoring of burn wound healing rely on medical staff visually observing the wound color and distribution of necrotic tissue, or using palpation on the back of the hand to judge skin temperature and the degree of redness and swelling around the wound. Due to subjective experience differences, the diagnostic results lack objective and unified standards. Collecting secretion samples with sterile cotton swabs for culture is time-consuming and cannot meet the need for immediate feedback. Using surgical instruments to cut granulation tissue for examination can cause secondary physical damage to the fragile wound tissue. It is difficult to continuously capture the dynamic coupling characteristics of microcirculation blood flow and physical morphological changes during the wound healing process. Summary of the Invention

[0004] To achieve the above objectives, the present invention employs the following technical solution: a method for real-time monitoring of burn wound healing process, comprising the following steps: S1: Collect oxygen saturation data and locate extreme points, interpolate and connect extreme points to construct an envelope, calculate the envelope difference to obtain the instantaneous difference sequence, calculate the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and construct a microcirculation blood flow feature set; S2: By segmenting the wound image using pixel color features, dividing the outer reference region and filtering the texture feature points within the region, calculating the displacement vectors at multiple time points and interpolating them, a global deformation vector field is constructed. S3: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation and migration vector field. Calculate the ratio of the vector field to the total physical displacement vector magnitude to generate the regeneration shrinkage ratio coefficient. S4: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, construct a state vector sequence, calculate the Euclidean distance between each pair of time vectors in the state vector sequence, filter coordinate positions with Euclidean distance less than the preset cutoff radius standard and perform binarization marking, and generate a recursive feature matrix. S5: Scan the vertical segments of the recursive feature matrix, calculate the proportion of vertical segment points to obtain the laminar flow index and the average length to obtain the capture time index, combine the amplitude fluctuation rate index in the microcirculation blood flow feature set and the regeneration contraction ratio coefficient to make a judgment, and output the wound healing assessment result.

[0005] As a further aspect of the present invention, the microcirculatory blood flow feature set specifically includes time-series data of wound tissue oxygen saturation, instantaneous difference sequence, and amplitude fluctuation rate index; the global deformation vector field specifically includes the displacement vector of texture feature points in the peripheral reference area and regional interpolation deformation distribution data; the regeneration shrinkage ratio coefficient specifically includes the magnitude value of the net proliferation migration vector field and the magnitude value of the total physical displacement vector; the recursive feature matrix specifically includes the state vector coordinate index of the multidimensional phase space and the binarized recursive feature value; and the wound healing assessment result specifically includes laminar flow index, capture time index, and pathological state classification identifier.

[0006] As a further aspect of the present invention, the step of obtaining the microcirculation blood flow feature set specifically includes: S101: Using a multispectral sensor, time-series data of oxygen saturation in wound tissue is collected. The data is traversed to locate the coordinates of multiple local maxima and multiple local minima. The cubic spline interpolation function is called to smoothly fit and connect the coordinates of multiple local maxima and multiple local minima respectively, construct the upper envelope and lower envelope, and output the oxygen saturation analysis results. S102: Using the oxygen saturation analysis results, the data sampling points of the upper and lower envelopes are synchronously aligned based on the timestamps, the difference data between the upper and lower envelopes at the same time is calculated, the time-varying fluctuation characteristics of the difference data are extracted and recombined according to the sampling time order to generate an instantaneous difference sequence. S103: For the instantaneous difference sequence, calculate the gradient of numerical change at adjacent time points point by point, accumulate and calculate the absolute value of the gradient and calculate the average value, obtain the amplitude fluctuation rate index, integrate the tissue oxygen saturation time series data, instantaneous difference sequence and amplitude fluctuation rate index, and generate a microcirculation blood flow feature set.

[0007] As a further aspect of the present invention, the step of obtaining the global deformation vector field specifically includes: S201: Acquire wound image data from continuous monitoring cycles, convert each frame to HSV color space and extract saturation component layers, set a preset threshold range matching the spectral characteristics of normal skin, perform binarization segmentation on the saturation component layers to obtain a binary mask image, perform morphological closing operation on the binary mask image, fill internal holes and smooth edge contours, extract the largest connected component as an effective mask, and generate an outer reference region. S202: For the outer reference area, filter the pixel coordinates of the corner response function values ​​that are greater than the preset feature threshold to obtain texture feature points, construct a multi-layer resolution image pyramid structure, search the matching coordinates with the smallest difference in local neighborhood gray level distribution between the texture feature points and the image at adjacent time points layer by layer from top to bottom, calculate the spatial vector difference between the matching coordinates and the original position, remove outlier vectors whose modulus exceeds the preset limit and aggregate the remaining results to generate displacement vector data. S203: Construct an irregular triangular mesh based on the coordinates of discrete feature points in the displacement vector data, establish a regular two-dimensional covering mesh based on the image pixel coordinate system, use radial basis function interpolation, calculate the estimated displacement of each node position in the two-dimensional covering mesh by weighting the displacement values ​​of the vertices of the irregular triangular mesh, map the estimated displacement to the corresponding mesh node coordinate index to establish a vector matrix, and generate a global deformation vector field.

[0008] As a further aspect of the present invention, the process of setting a preset threshold range that matches the spectral characteristics of normal skin specifically involves: Collect healthy skin sample images under the same lighting conditions as the wound image data, extract the pixel values ​​of the saturation channel in HSV space, calculate the expected value and standard deviation of all saturation channel pixel values, set the sum of the expected value and 3 times the standard deviation as the upper limit of the threshold, and set the difference between the expected value and 3 times the standard deviation as the lower limit of the threshold to generate the preset threshold range. The process of setting the preset feature threshold is as follows: Calculate the horizontal and vertical gradient components of the pixels within the outer reference region, construct the second-order moment matrix and calculate the corresponding corner response function values, count all corner response function values, calculate the segmentation value when the inter-class variance is maximized using the maximum inter-class variance method, and set it as the preset feature threshold.

[0009] As a further aspect of the present invention, the step of obtaining the regeneration shrinkage specific gravity coefficient specifically includes: S301: Calculate the derivatives of pixel grayscale in the horizontal and vertical directions for wound image data, construct a texture gradient matrix and locate grayscale abrupt change regions, perform curve fitting on the pixel set to generate physical edge lines, track the coordinate changes of discrete sampling points on the physical edge lines in continuous time series, calculate the spatial vector difference between the current coordinates and the initial coordinates, and generate the total physical displacement vector. S302: Read the coordinates of the discrete sampling points associated with the total physical displacement vector, project the coordinates onto the spatial grid system of the global deformation vector field, calculate the background reference displacement vector at each projection point based on the grid node values, subtract the corresponding background reference displacement vector from the total physical displacement vector point by point, remove the background deformation interference component, and generate the net proliferation migration vector field. S303: Calculate the magnitude of each vector element in the net proliferation migration vector field to obtain the net proliferation magnitude sequence, simultaneously calculate the magnitude of the corresponding vector element in the total physical displacement vector to obtain the total displacement magnitude sequence, perform point-to-point division to calculate the ratio of the net proliferation magnitude to the total displacement magnitude, calculate the arithmetic mean of all the ratio values, and generate the regeneration shrinkage ratio coefficient.

[0010] As a further aspect of the present invention, the process of locating the grayscale abrupt change region specifically includes: Extract the horizontal and vertical derivatives from the texture gradient matrix, calculate the square root of their sum to obtain the gradient magnitude, and calculate the arctangent of their ratio to obtain the gradient direction. Compare the gradient magnitude of the current pixel with that of its two adjacent pixels along the gradient direction. If the current pixel is not a local maximum, set the gradient magnitude to zero and perform non-maximum suppression. Statistically analyze the gradient magnitude distribution data after processing, calculate the segmentation value when the inter-class variance is maximized using the maximum inter-class variance method, set it as the adaptive edge segmentation threshold, and filter pixels with gradient magnitudes greater than the adaptive edge segmentation threshold to establish the gray-scale abrupt change region.

[0011] As a further aspect of the present invention, the step of obtaining the recursive feature matrix specifically includes: S401: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, set the embedding dimension and time delay parameters, perform time series reconstruction processing on the tissue oxygen saturation time series data according to the time delay parameters, convert the one-dimensional time series data into discrete vectors in a multi-dimensional spatial coordinate system, arrange all discrete vectors according to the original sampling time order, and generate a state vector sequence. S402: Perform a full permutation traversal operation on the state vector sequence, extract the vector coordinates corresponding to any two time points, calculate the sum of squares of the differences between the two vectors in each dimension of the multidimensional space and perform a square root operation, obtain the Euclidean distance value of the vector pair at the corresponding time point, construct a two-dimensional square matrix with the time point as the row and column index, fill the Euclidean distance value into the corresponding coordinate position of the two-dimensional square matrix, and generate a vector spacing matrix. S403: Obtain the preset cutoff radius standard, traverse the Euclidean distance values ​​stored in the vector spacing matrix, compare the Euclidean distance values ​​with the preset cutoff radius standard item by item, filter the coordinate positions where the Euclidean distance value is less than the cutoff radius standard and perform binarization marking, and generate a recursive feature matrix.

[0012] As a further aspect of the present invention, the process of obtaining the preset cutoff radius standard is specifically as follows: The maximum Euclidean distance value is obtained by traversing the vector spacing matrix. Multiple incremental test radii are set with the value from zero to the maximum Euclidean distance value as the interval. The number of elements in the vector spacing matrix with values ​​less than each test radius is counted, and the proportion of the elements to the total number of non-zero elements in the matrix is ​​calculated. An associated integral sequence is constructed. A double logarithmic coordinate transformation is performed on the associated integral sequence and the corresponding test radius. The local slope of the transformed curve is calculated. A linear scaling region with a stable local slope value is identified. The test radius corresponding to the right endpoint of the linear scaling region is extracted and set as the preset cutoff radius standard.

[0013] As a further aspect of the present invention, the steps for obtaining the wound healing assessment results are specifically as follows: S501: Scan the recursive feature matrix to identify vertical line segment structures parallel to the main diagonal, calculate the numerical ratio of the total number of recursive points in the vertical line segment structure to the total number of recursive points in the whole matrix to obtain the laminar flow index, calculate the arithmetic mean of the lengths of all vertical line segment structures to obtain the capture time index, and generate a set of nonlinear dynamic parameters. S502: The amplitude fluctuation rate index is extracted by calling the microcirculation blood flow feature set, the regeneration contraction ratio coefficient is read, and multi-dimensional feature splicing is performed with the nonlinear dynamic parameter set. The spatial distance between the spliced ​​feature vector and the preset pathological benchmark vector is calculated. The current category label is determined based on the minimum distance, the pathological state classification identifier is obtained, and the wound healing assessment result is generated.

[0014] As a further aspect of the present invention, the process of setting the preset pathological baseline vector is specifically as follows: Historical wound monitoring samples with pathological stage labels are obtained. Laminar flow index, capture time index, amplitude fluctuation rate index, and regeneration shrinkage ratio coefficient are extracted for each sample. The values ​​of the above indicators are normalized and spliced ​​in a fixed order to construct a historical sample feature vector. The historical sample feature vector is divided into different pathological state subsets according to the pathological stage label. The arithmetic mean of all vectors in each pathological state subset is calculated in each dimension to generate the cluster center vector of each pathological stage, which is set as the preset pathological baseline vector.

[0015] A real-time monitoring system for the healing process of burn wounds, including: Signal feature extraction module: Collects oxygen saturation data and locates extreme points, interpolates and connects extreme points to construct an envelope, calculates the envelope difference to obtain the instantaneous difference sequence, calculates the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and constructs a microcirculation blood flow feature set; Deformation field dynamic tracking module: By segmenting the wound image by pixel color features, dividing the outer reference area and filtering the texture feature points within the area, calculating the displacement vector between multiple time points and interpolating it, a global deformation vector field is constructed. Regeneration weight analysis module: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation migration vector field, calculate the ratio with the total physical displacement vector magnitude, and generate the regeneration shrinkage weight coefficient; The time-series recursive analysis module calls the microcirculation blood flow feature set, extracts tissue oxygen saturation time-series data, constructs a state vector sequence, calculates the Euclidean distance between each pair of time vectors in the state vector sequence, filters out coordinate positions with Euclidean distances less than the preset cutoff radius standard and performs binarization marking, and generates a recursive feature matrix. Healing process determination module: Scans the vertical segments of the recursive feature matrix, calculates the proportion of vertical segment points to obtain laminar flow index and average length to obtain capture time index, combines the amplitude fluctuation rate index in the microcirculation blood flow feature set and the regeneration contraction ratio coefficient to make a determination, and outputs the wound healing assessment result.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, spectral data is collected and a microcirculatory blood flow feature set is constructed. By combining wound texture gradient calculation and deformation field mapping, background motion interference is eliminated to accurately quantify net proliferation and migration behavior. Nonlinear dynamics methods are used to map time-series signals to a multidimensional phase space and construct a recursive feature matrix. Laminar flow and capture time indicators in state evolution are identified from complex fluctuations. Physiological blood flow parameters and morphological contraction coefficients are fused to perform multidimensional state mapping judgment, realizing non-contact objective quantitative analysis of the healing process. This avoids the infection risk caused by physical contact and eliminates the subjective bias of human experience judgment. A multidimensional pathological grading standard is established to accurately reflect the dynamic trend of tissue remodeling. Attached Figure Description

[0017] 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.

[0018] 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; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[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 a method for real-time monitoring of the healing process of burn wounds, comprising the following steps: S1: Collect oxygen saturation data and locate extreme points, interpolate and connect extreme points to construct an envelope, calculate the envelope difference to obtain the instantaneous difference sequence, calculate the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and construct a microcirculation blood flow feature set; S2: By segmenting the wound image using pixel color features, dividing the outer reference region and filtering the texture feature points within the region, calculating the displacement vectors at multiple time points and interpolating them, a global deformation vector field is constructed. S3: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation and migration vector field. Calculate the ratio of the vector field to the total physical displacement vector magnitude to generate the regeneration shrinkage ratio coefficient. S4: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, construct a state vector sequence, calculate the Euclidean distance between each pair of time vectors in the state vector sequence, filter coordinate positions with Euclidean distance less than the preset cutoff radius standard and perform binarization marking, and generate a recursive feature matrix. S5: Scan the vertical line segments of the recursive feature matrix, calculate the proportion of vertical line segment points to obtain laminar flow index and average length to obtain capture time index, combine the amplitude fluctuation rate index and regeneration contraction ratio coefficient in the microcirculation blood flow feature set to make a judgment, and output the wound healing assessment result.

[0022] The microcirculation blood flow feature set specifically includes time-series data of wound tissue oxygen saturation, instantaneous difference sequence, and amplitude fluctuation rate index; the global deformation vector field specifically includes the displacement vector of texture feature points in the peripheral reference area and regional interpolation deformation distribution data; the regeneration shrinkage ratio coefficient specifically includes the magnitude value of the net proliferation migration vector field and the magnitude value of the total physical displacement vector; the recursive feature matrix specifically includes the state vector coordinate index of the multidimensional phase space and the binary recursive feature value; and the wound healing assessment results specifically include laminar flow index, capture time index, and pathological state classification identifier.

[0023] Please see Figure 2 The specific steps for obtaining the microcirculatory blood flow feature set are as follows: S101: Using a multispectral sensor, time-series data of oxygen saturation in wound tissue is collected. The data is traversed to locate the coordinates of multiple local maxima and multiple local minima. The cubic spline interpolation function is called to smoothly fit and connect the coordinates of multiple local maxima and multiple local minima respectively, construct the upper envelope and lower envelope, and output the oxygen saturation analysis results. First, the multispectral imaging component is activated, and the spectral acquisition band is set to cover the visible to near-infrared light range, for example, from 500 nm to 900 nm. The burn wound area is continuously scanned at a preset sampling frequency, such as 50 frames per second, to obtain a sequence of oxygen saturation values ​​at numerous discrete time points. Then, an extreme point localization operation is performed. The entire acquired oxygen saturation time-series data is traversed, and a sliding window comparison method is used to compare the current data point with its multiple adjacent data points. If the oxygen saturation value of the current data point is strictly greater than the values ​​of all data points within its adjacent time window, the point is determined to be a local maximum, and its corresponding timestamp and numerical coordinates are recorded. Conversely, if the value of the current data point is strictly less than its adjacent data points, it is determined to be a local minimum. For example, if, during a certain monitoring period, the oxygen saturation values ​​at the 10th, 25th, and 40th seconds are 98, 99, and 97 respectively, and all are higher than their neighboring values, then this point is marked as a local maximum. Next, a cubic spline interpolation function is invoked to smoothly fit the selected discrete extrema. This interpolation process constructs a piecewise cubic polynomial, ensuring the continuity of the first and second derivatives at each extremum point. This allows for the connection of all local maxima to generate the upper envelope, and the connection of all local minima to generate the lower envelope. This operation not only fills the time gaps between extrema points but also suppresses the interference of high-frequency acquisition noise through smoothing, ultimately outputting oxygen saturation analysis results that reflect the fluctuation range of microcirculatory blood perfusion.

[0024] Table 1. Data Collection and Extreme Point Indication of Oxygen Saturation in Wounds Time point number Sampling time (seconds) Measured oxygen saturation Extreme value type marker Upper envelope fitted value Lower envelope fitted value 1 0.5 88 none 92.5 88.0 20 10.0 98 Local maxima 98.0 85.0 50 25.0 82 Local Minimum 97.5 82.0 80 40.0 99 Local maxima 99.0 83.5 As shown in Table 1, the table displays the measured oxygen saturation data at some sampling times and the fitted envelope values ​​generated after extreme value determination. Through the cubic spline interpolation algorithm, even at sampling times that are not extreme points (such as number 1), the corresponding envelope boundary values ​​can be calculated based on the trend of the extreme points before and after, thus establishing the dynamic range of fluctuation.

[0025] S102: Using the oxygen saturation analysis results, the data sampling points of the upper and lower envelopes are synchronously aligned based on the timestamps. The difference data between the upper and lower envelopes at the same time is calculated. The time-varying fluctuation characteristics of the difference data are extracted and recombined according to the sampling time order to generate an instantaneous difference sequence. The upper and lower envelope data structures are read, and based on a unified original sampling timestamp as a reference index, it is checked whether there are corresponding data points for each discrete sampling moment on the two curves. If there is a slight time deviation caused by interpolation, linear interpolation is used for fine-tuning to force alignment of the data sampling points of the upper and lower envelopes at the same moment, ensuring that subsequent calculations are performed in a strictly consistent time dimension. Subsequently, a difference calculation operation is performed. For each synchronized sampling moment, the value of the upper envelope at that moment is subtracted from the value of the lower envelope at the same moment to obtain the amplitude difference between the two. This difference data directly reflects the dynamic range of the pulsation amplitude or regulatory capacity of the microcirculation blood perfusion in the wound tissue at a specific time point. For example, at the 10th second, the upper envelope value is 98 and the lower envelope value is 85, then the difference data calculation result at that moment is 13. The above subtraction operation is performed sequentially on all sampling points throughout the entire monitoring period to extract a series of difference values ​​reflecting time-varying fluctuation characteristics. These values ​​are then rearranged and recombined strictly according to the original sampling time sequence to generate a continuous instantaneous difference sequence. This sequence eliminates the general DC component trend and focuses on quantifying the transient regulatory capacity of microcirculation and the activity of vasomotor activity.

[0026] S103: For the instantaneous difference sequence, calculate the gradient of numerical change between adjacent time points point by point, accumulate and calculate the absolute value of the gradient and calculate the average value, obtain the amplitude fluctuation rate index, integrate the tissue oxygen saturation time series data, instantaneous difference sequence and amplitude fluctuation rate index to generate a microcirculation blood flow feature set; The algorithm iterates through each data point in the instantaneous difference sequence, calculating the numerical change gradient between adjacent time points. This is done by subtracting the instantaneous difference of the previous time point from the current instantaneous difference and dividing by the time interval between the two sampling points to obtain the rate of change at that moment. To comprehensively assess the severity of fluctuations without being affected by the direction of change, the absolute values ​​of all calculated gradient values ​​are then calculated, unifying negative attenuation and positive enhancement into magnitude. Subsequently, the absolute values ​​of all gradient values ​​within a set monitoring window are summed, and the sum is divided by the total number of data points within that window to calculate the arithmetic mean. This mean is defined as the amplitude fluctuation rate index. A larger index value indicates more severe blood flow fluctuations in the wound, indicating an unstable microcirculation; a smaller value indicates smoother blood flow fluctuations. For example, if the calculated absolute value sequence of gradient values ​​is 0.5, 0.4, and 0.6, the sum is 1.5, and the average is 0.5, which is the amplitude fluctuation rate index. Finally, a feature integration operation is performed to package and structure the original wound tissue oxygen saturation time series data, the calculated instantaneous difference sequence, and the finally obtained amplitude fluctuation rate index into a microcirculation blood flow feature set containing multidimensional physiological information, providing underlying data support for subsequent pathological state analysis.

[0027] Please see Figure 3 The specific steps for obtaining the global deformation vector field are as follows: S201: Acquire wound image data from continuous monitoring cycles, convert each frame to HSV color space and extract saturation component layers, set a preset threshold range matching the spectral characteristics of normal skin, perform binarization segmentation on the saturation component layers to obtain a binary mask image, perform morphological closing operation on the binary mask image, fill internal holes and smooth edge contours, extract the largest connected component as an effective mask, and generate an outer reference region. The acquired RGB format wound images are converted to the HSV color space frame by frame, as the HSV space can better separate luminance and chromaticity information. Within the HSV space, the saturation component layer (S channel) is extracted separately, as this channel is highly sensitive to the blood filling of skin tissue. Next, a preset threshold range is set. This process involves acquiring healthy skin sample images under the same lighting conditions as the current wound image, extracting their S channel pixel values, and calculating the expected value (mean) and standard deviation of these pixel values. For example, if the expected value of the healthy skin S channel is 100 and the standard deviation is 10, the upper threshold is set to the sum of the expected value and three times the standard deviation (i.e., 130), and the lower threshold is set to the difference between the expected value and three times the standard deviation (i.e., 70). Using this generated preset threshold range [70, 130], binarization segmentation is performed on the saturation component layer of the wound image: pixels with values ​​within this range are marked as foreground (e.g., white), and pixels outside the range are marked as background (e.g., black), thus obtaining a binary mask image. To eliminate noise interference, a morphological closing operation is performed on the binary mask image, namely dilation followed by erosion, to fill the tiny voids inside the mask and smooth the edge contours. Finally, a connected component analysis algorithm is run to identify all connected regions in the image, and the connected component with the largest area is extracted as the effective mask. The area covered by this effective mask is defined as the outer reference region, which represents the relatively healthy skin tissue around the wound and is used for subsequent background motion estimation.

[0028] Table 2. Threshold setting parameters for normal skin spectral characteristics Parameter name Mathematical expectation Standard deviation value Calculate the multiplier Threshold lower limit results Threshold upper limit result S-channel pixel value 100 10 3 70 130 As shown in Table 2, the threshold range determined by statistical methods can adaptively cover the spectral characteristics of normal skin. The calculation logic is as follows: upper threshold = 100 + (3 × 10) = 130; lower threshold = 100 - (3 × 10) = 70. This range ensures high robustness in the extraction of the peripheral reference area and eliminates interference caused by changes in illumination.

[0029] S202: For the outer reference area, filter the pixel coordinates of the corner response function values ​​that are greater than the preset feature threshold to obtain texture feature points, construct a multi-layer resolution image pyramid structure, search the matching coordinates with the smallest difference in local neighborhood gray level distribution between the texture feature points and the image at adjacent time points layer by layer from top to bottom, calculate the spatial vector difference between the matching coordinates and the original position, remove outlier vectors whose modulus exceeds the preset limit and aggregate the remaining results to generate displacement vector data. For each pixel within the region, its horizontal and vertical gradient components are calculated. These two components are used to construct a second-order moment matrix, and the corner response function value (e.g., Harris response value) is calculated based on this matrix. The corner response function values ​​of all pixels within the region are statistically analyzed, and a segmentation value is automatically calculated using the Otsu's algorithm (maximum inter-class variance method), which is then set as a preset feature threshold. Pixel coordinates with corner response function values ​​greater than this preset feature threshold are selected and defined as texture feature points. These points are typically located at blood vessel intersections or areas with prominent skin textures, exhibiting good trackability. To improve the adaptability to large displacements during tracking, a multi-level image pyramid structure is constructed, and the original image is downsampled multiple times. During the tracking phase, optical flow or block matching algorithms are used to search layer by layer from the top to the bottom of the pyramid. The coordinates with the smallest difference in local neighborhood grayscale distribution (e.g., smallest sum of squared errors) between adjacent time-series images and the current texture feature point are identified as the matching coordinates. The spatial vector difference between the matching coordinates and the original position is calculated to obtain the initial displacement vector. To eliminate false matches, the magnitude of all vectors is calculated, outlier vectors whose magnitude exceeds a preset limit (e.g., more than 5% of the total image size) are removed, and the remaining valid results are aggregated to generate highly reliable displacement vector data. This data accurately describes the physical displacement of the skin background of the wound caused by breathing or body movement.

[0030] S203: Construct an irregular triangular mesh based on the coordinates of discrete feature points in the displacement vector data, establish a regular two-dimensional covering mesh based on the image pixel coordinate system, use radial basis function interpolation, calculate the estimated displacement of each node position in the two-dimensional covering mesh by weighting the displacement values ​​of the vertices of the irregular triangular mesh, map the estimated displacement to the corresponding mesh node coordinate index to establish a vector matrix, and generate a global deformation vector field. Using discrete texture feature point coordinates as vertices, an irregular triangular mesh (such as a Delaunay triangular mesh) is constructed, dividing the outer reference region into several non-overlapping triangular units. Simultaneously, a regularized two-dimensional overlay mesh is established based on the image pixel coordinate system, covering the entire wound surface and its surrounding area. Next, region interpolation is performed using the radial basis function (RBF) interpolation algorithm. Based on the known displacement values ​​of each vertex of the irregular triangular mesh, the estimated displacement of each node in the two-dimensional overlay mesh is calculated using weighted averages. This interpolation logic assumes that the closer a mesh node is to a feature point in space, the greater the influence of that feature point on its motion state. All mesh nodes are traversed, and the background displacement vector (including horizontal and vertical components) of each node at the current moment is calculated using the above interpolation algorithm. These estimated displacements are then mapped to the corresponding mesh node coordinate indices, establishing a dense vector matrix. This matrix is ​​the global deformation vector field, which details the displacement field distribution of each pixel position in the image background due to non-wound contraction factors (such as breathing and body micro-movements), providing a spatial reference for subsequent background interference removal.

[0031] Please see Figure 4 The specific steps for obtaining the regeneration shrinkage density coefficient are as follows: S301: Calculate the derivatives of pixel grayscale in the horizontal and vertical directions for wound image data, construct a texture gradient matrix and locate grayscale abrupt change regions, perform curve fitting on the pixel set to generate physical edge lines, track the coordinate changes of discrete sampling points on the physical edge lines in continuous time series, calculate the spatial vector difference between the current coordinates and the initial coordinates, and generate the total physical displacement vector. Differentiate the image pixel grayscale values, calculating the horizontal and vertical derivatives to construct a texture gradient matrix. Based on this matrix, locate regions of abrupt grayscale changes. Specifically: extract the horizontal and vertical derivatives, calculate the square root of their sum to obtain the gradient magnitude, and calculate the arctangent of the ratio to obtain the gradient direction. Compare the gradient magnitudes of adjacent pixels along the gradient direction; if the current point is not a local maximum, set it to zero (non-maximum suppression). Then, use the Otsu's method to calculate an adaptive edge segmentation threshold, filtering out pixels with gradient magnitudes greater than this threshold. These points constitute regions of abrupt grayscale changes, i.e., potential wound edges. Next, perform curve fitting (e.g., B-spline curve fitting) on ​​these discrete sets of edge pixels to generate smooth, closed physical edge lines. In continuous temporal images, track the coordinate changes of multiple discrete sampling points (e.g., keypoints sampled at equal arc lengths) on this physical edge line. Calculate the spatial vector difference between the coordinates of each sampling point at the current time and its coordinates at the initial time. For example, if the initial coordinates of an edge point are (100, 100) and the current coordinates are (95, 95), then its spatial vector difference is (-5, -5). The vector differences of all sampling points along the edge line are summarized to generate a total physical displacement vector, which includes the contraction movement of the wound itself and the overall movement of the background.

[0032] S302: Read the coordinates of discrete sampling points associated with the total physical displacement vector, project the coordinates onto the spatial grid system of the global deformation vector field, calculate the background reference displacement vector at each projection point based on the grid node values, subtract the corresponding background reference displacement vector from the total physical displacement vector point by point, remove the background deformation interference component, and generate the net proliferation migration vector field. Read the current coordinates of each discrete sampling point in the total physical displacement vector calculated above. Project these coordinates onto the spatial grid system of the global deformation vector field constructed in step S203. For each projection point, based on its position in the grid, use bilinear interpolation to resolve the background reference displacement vector at that point from the estimated displacement of the surrounding grid nodes. This background reference displacement vector represents the displacement that the point should have if it only moves with the background of healthy skin (such as breathing) without wound contraction. Next, perform the core denoising operation: subtract the corresponding background reference displacement vector from the total physical displacement vector point by point. Vector subtraction follows the parallelogram law, and the values ​​are subtracted on the horizontal and vertical components respectively. For example, if the total physical displacement vector of an edge point is (-5, -5), and the background reference displacement vector corresponding to that position is (-1, -1) (representing the displacement caused by breathing), then perform the operation: (-5 - (-1), -5 - (-1)) = (-4, -4). The result (-4, -4) is the net proliferation and migration vector after removing background deformation interference components. It purely reflects the centripetal contraction or epithelial creeping behavior of the wound edge tissue due to the healing mechanism. Perform this operation on all edge sampling points to finally generate the net proliferation and migration vector field.

[0033] S303: Calculate the magnitude of each vector element in the net proliferation migration vector field to obtain the net proliferation magnitude sequence, simultaneously calculate the magnitude of the corresponding vector element in the total physical displacement vector to obtain the total displacement magnitude sequence, perform point-to-point division to calculate the ratio of the net proliferation magnitude to the total displacement magnitude, calculate the arithmetic mean of all the ratio values, and generate the regeneration shrinkage ratio coefficient. The algorithm iterates through each element of the net proliferation migration vector field, calculating its magnitude (i.e., vector length) using the Pythagorean theorem to obtain the net proliferation magnitude sequence. Simultaneously, iterates through the corresponding elements of the total physical displacement vector, calculating their magnitudes to obtain the total displacement magnitude sequence. Next, a point-to-point division operation is performed. For each sampling point on the edge line, its net proliferation magnitude is divided by its total displacement magnitude to obtain a percentage. This percentage reflects the proportion of the total movement at that point that is truly contributed by healing contraction. For example, if the net proliferation magnitude is 5.65 (corresponding to vector (-4, -4)) and the total displacement magnitude is 7.07 (corresponding to vector (-5, -5)), then the percentage at that point is 5.65 / 7.07 ≈ 0.8. Finally, the percentage values ​​of all sampling points on the edge line are statistically analyzed, and their arithmetic mean is calculated. This average result is the regeneration contraction proportion coefficient. This coefficient is a value between 0 and 1. The closer the value is to 1, the more dominant the wound displacement is in terms of healing contraction, with less background interference. The lower the value, the more significant the background motion interference or the weaker the wound contraction. This coefficient provides a normalized quantitative indicator for assessing the effectiveness of wound contraction function.

[0034] Please see Figure 5 The specific steps for obtaining the recursive feature matrix are as follows: S401: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, set the embedding dimension and time delay parameters, perform time series reconstruction processing on tissue oxygen saturation time series data according to the time delay parameters, convert one-dimensional time series data into discrete vectors in a multi-dimensional spatial coordinate system, arrange all discrete vectors according to the original sampling time order, and generate a state vector sequence. First, two key parameters are defined: the embedding dimension (m) and the time delay parameter (τ). The embedding dimension determines the dimension of the reconstruction space, and the time delay parameter determines the interval between sampling points. Based on these two parameters, temporal reconstruction processing is performed on one-dimensional tissue oxygen saturation time-series data. Specifically, for each time point t in the time-series data, a multi-dimensional vector is constructed. The components of this vector consist of data at time t, data at time t+τ, and data up to time t+(m-1)τ. For example, if the embedding dimension m=3, the time delay τ=2, and the time-series data is [x1, x2, x3, x4, x5, x6...], then the first generated state vector is [x1, x3, x5], the second state vector is [x2, x4, x6], and so on. In this way, the oxygen saturation signal, which originally varies on a one-dimensional time axis, is converted into a series of discrete vectors in a multi-dimensional spatial coordinate system. All the generated discrete vectors are arranged in the order of the original sampling times to generate a sequence of state vectors. This process aims to reveal the high-dimensional dynamic attractor structure hidden behind one-dimensional time series, which is a prerequisite for analyzing the chaotic characteristics of nonlinear physiological signals.

[0035] S402: Perform a full permutation traversal operation on the state vector sequence, extract the vector coordinates corresponding to any two time points, calculate the sum of squares of the differences between the two vectors in each dimension of the multidimensional space and perform the square root operation, obtain the Euclidean distance value of the vector pair at the corresponding time point, construct a two-dimensional square matrix with the time point as the row and column index, fill the Euclidean distance value into the corresponding coordinate position of the two-dimensional square matrix, and generate the vector spacing matrix. Extract the vector coordinates (Vi and Vj) corresponding to any two time points (time i and time j) in the sequence. Calculate the Euclidean distance between these two vectors in multidimensional space. The specific calculation logic is as follows: First, calculate the numerical difference between the two vectors in each dimension component. Square these differences respectively. Then, sum the squared differences of all dimensions. Finally, perform a square root operation (i.e., arithmetic square root) on the sum to obtain the Euclidean distance value of the vector pair at the corresponding time point. For example, if vectors Vi=[1, 2] and Vj=[4, 6], then the difference in each dimension is [3, 4], which squares to [9, 16], and the sum is 25. The square root yields an Euclidean distance of 5. Construct a two-dimensional square matrix (N×N matrix, where N is the total number of vectors) with time points as row and column indices. Fill the calculated Euclidean distance value into the corresponding coordinate positions (i, j) in this two-dimensional square matrix. Iterate through all possible combinations of i and j, filling the entire matrix to generate the vector distance matrix. This matrix fully records the distance relationships between all state trajectory points in the phase space, reflecting the degree of mutual proximity of states during the evolution process.

[0036] S403: Obtain the preset cutoff radius standard, traverse the Euclidean distance values ​​stored in the vector spacing matrix, compare the Euclidean distance values ​​with the preset cutoff radius standard item by item, filter the coordinate positions where the Euclidean distance value is less than the cutoff radius standard and perform binarization marking, and generate a recursive feature matrix. The maximum Euclidean distance value is obtained by traversing the vector spacing matrix. Multiple incremental test radii are set with the range from 0 to the maximum distance. The number of elements in the matrix whose values ​​are less than each test radius is counted, and the proportion of these elements to the total number of non-zero elements is calculated to construct an associated integral sequence. A double logarithmic coordinate transformation is performed on this sequence and the test radii to calculate the local slope of the curve. Linear scaling regions with stable slopes are identified, and the test radius corresponding to the right endpoint of these regions is extracted as a preset cutoff radius standard. After obtaining this standard, each Euclidean distance value stored in the vector spacing matrix is ​​traversed again, and each value is compared: if the Euclidean distance value at a certain position is less than the preset cutoff radius standard, it indicates that the two positions are in similar states (recursion), and the position is marked as 1 (black / recursive); otherwise, if it is greater than the standard, it is marked as 0 (white / no recursion). Through this binarization process, the distance matrix is ​​converted into a recursive feature matrix (i.e., a recursive graph). This matrix visually demonstrates the deterministic, periodic, and stable nonlinear dynamic characteristics of wound microcirculation through a visualized texture structure (such as diagonals, vertical lines, and block structures).

[0037] Please see Figure 6 The specific steps for obtaining the wound healing assessment results are as follows: S501: Scan the recursive feature matrix to identify vertical line segment structures parallel to the main diagonal, calculate the numerical ratio of the total number of recursive points in the vertical line segment structure to the total number of recursive points in the whole matrix to obtain the laminar flow index, calculate the arithmetic mean of the lengths of all vertical line segment structures to obtain the capture time index, and generate a set of nonlinear dynamic parameters. Identify vertical line segments parallel to the main diagonal in the matrix. In the recursion graph, vertical segments indicate a state of laminarity where the state remains largely unchanged over a period of time. Scan each column and count all vertical segments consisting of consecutive "1"s. Calculate the laminarity index by summing the total number of recursive points (i.e., "1") contained in all vertical segments and dividing by the total number of recursive points in the entire matrix to obtain a proportional value. The higher this proportion, the more stable the state. Simultaneously, calculate the capture time index, which is the arithmetic mean of the lengths of all identified vertical segments. This index reflects the average residence time in a specific state. For example, if three vertical segments with lengths of 3, 4, and 5 are identified in the matrix, the average length is 4. This capture time index quantifies the ability of the wound microcirculation to recover or maintain a steady state after being disturbed. Package the calculated laminarity index and capture time index to generate a set of nonlinear dynamic parameters. These parameters reveal the changes in microscopic stability during the wound healing process from the perspective of chaos theory.

[0038] S502: Call the microcirculation blood flow feature set to extract the amplitude fluctuation rate index, read the regeneration contraction ratio coefficient, perform multi-dimensional feature splicing with the nonlinear dynamic parameter set, calculate the spatial distance between the spliced ​​feature vector and the preset pathological baseline vector, determine the current category label based on the minimum distance, obtain the pathological state classification label, and generate wound healing assessment results; First, the microcirculation blood flow feature set is accessed, and the amplitude fluctuation rate index is extracted. The regeneration contraction ratio coefficient is read, and combined with the nonlinear dynamic parameter set (laminar flow index, capture time index). The feature values ​​of these four dimensions are normalized (e.g., mapped to the 0-1 interval) and concatenated in a fixed order to form a comprehensive feature vector of the current wound. Next, the spatial distance between this concatenated feature vector and a preset pathological baseline vector is calculated. The logic for obtaining the preset pathological baseline vector is as follows: based on historical samples, it is classified according to pathological stages (e.g., inflammation, proliferative, remodeling, infection), and the cluster center (arithmetic mean) of the feature vectors under each category is calculated. The Euclidean distance between the current feature vector and each pathological baseline vector (cluster center) is calculated, and the category to which the current wound belongs is determined according to the minimum distance principle. For example, if the current vector is closest to the center of the "proliferative" stage, it is determined that the current wound is in the proliferative stage. Finally, the corresponding pathological state classification identifier is obtained and output along with the specific laminar flow index and capture time index to generate a complete wound healing assessment result.

[0039] Table 3. Sample Table of Wound Pathological Status Judgment Logic and Benchmark Vector Pathological state classification Laminar flow index benchmark value Capture Time Metric Benchmark Amplitude decay rate reference value Regeneration shrinkage coefficient benchmark value Inflammatory response period 0.2 2.5 0.8 0.3 tissue proliferative phase 0.6 5.0 0.4 0.7 Organizational restructuring period 0.8 8.0 0.2 0.9 As shown in Table 3, typical physical characteristic values ​​for different healing stages are preset as the original benchmarks. During the judgment process, the benchmark values ​​in Table 3 and the various indicators of the current monitoring data are uniformly mapped to the dimensionless [0, 1] interval using the preset data extreme value range, so as to eliminate the excessive influence of the capture time indicator (with a large numerical magnitude) on the Euclidean distance calculation weight.

[0040] Please see Figure 7 A real-time monitoring system for the healing process of burn wounds, including: Signal feature extraction module: Collects oxygen saturation data and locates extreme points, interpolates and connects extreme points to construct an envelope, calculates the envelope difference to obtain the instantaneous difference sequence, calculates the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and constructs a microcirculation blood flow feature set; Deformation field dynamic tracking module: By segmenting the wound image by pixel color features, dividing the outer reference area and filtering the texture feature points within the area, calculating the displacement vector between multiple time points and interpolating it, a global deformation vector field is constructed. Regeneration weight analysis module: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation migration vector field, calculate the ratio with the total physical displacement vector magnitude, and generate the regeneration shrinkage weight coefficient; The time-series recursive analysis module calls the microcirculation blood flow feature set, extracts tissue oxygen saturation time-series data, constructs a state vector sequence, calculates the Euclidean distance between each pair of time vectors in the state vector sequence, filters out coordinate positions with Euclidean distance less than the preset cutoff radius standard and performs binarization marking, and generates a recursive feature matrix. Healing process determination module: Scan the vertical line segments of the recursive feature matrix, calculate the proportion of vertical line segment points to obtain laminar flow index and average length to obtain capture time index, combine the amplitude fluctuation rate index and regeneration contraction ratio coefficient in the microcirculation blood flow feature set to make a determination, and output the wound healing assessment result.

[0041] The above are merely specific embodiments 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 method for real-time monitoring of burn wound healing process, characterized in that, Includes the following steps: S1: Collect oxygen saturation data and locate extreme points, interpolate and connect extreme points to construct an envelope, calculate the envelope difference to obtain the instantaneous difference sequence, calculate the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and construct a microcirculation blood flow feature set; S2: By segmenting the wound image using pixel color features, dividing the outer reference region and filtering the texture feature points within the region, calculating the displacement vectors at multiple time points and interpolating them, a global deformation vector field is constructed. S3: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation and migration vector field. Calculate the ratio of the vector field to the total physical displacement vector magnitude to generate the regeneration shrinkage ratio coefficient. S4: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, construct a state vector sequence, calculate the Euclidean distance between each pair of time vectors in the state vector sequence, filter coordinate positions with Euclidean distance less than the preset cutoff radius standard and perform binarization marking, and generate a recursive feature matrix. S5: Scan the vertical segments of the recursive feature matrix, calculate the proportion of vertical segment points to obtain the laminar flow index and the average length to obtain the capture time index, combine the amplitude fluctuation rate index in the microcirculation blood flow feature set and the regeneration contraction ratio coefficient to make a judgment, and output the wound healing assessment result.

2. The method for real-time monitoring of burn wound healing process according to claim 1, characterized in that, The microcirculation blood flow feature set specifically includes time-series data of wound tissue oxygen saturation, instantaneous difference sequence, and amplitude fluctuation rate index; the global deformation vector field specifically includes the displacement vector of texture feature points in the peripheral reference area and regional interpolated deformation distribution data; the regeneration shrinkage ratio coefficient specifically includes the magnitude value of the net proliferation migration vector field and the magnitude value of the total physical displacement vector; the recursive feature matrix specifically includes the state vector coordinate index of the multidimensional phase space and the binarized recursive feature value; and the wound healing assessment result specifically includes laminar flow index, capture time index, and pathological state classification identifier.

3. The method for real-time monitoring of burn wound healing process according to claim 1, characterized in that, The specific steps for obtaining the microcirculation blood flow feature set are as follows: S101: Using a multispectral sensor, time-series data of oxygen saturation in wound tissue is collected. The data is traversed to locate the coordinates of multiple local maxima and multiple local minima. The cubic spline interpolation function is called to smoothly fit and connect the coordinates of multiple local maxima and multiple local minima respectively, construct the upper envelope and lower envelope, and output the oxygen saturation analysis results. S102: Using the oxygen saturation analysis results, the data sampling points of the upper and lower envelopes are synchronously aligned based on the timestamps, the difference data between the upper and lower envelopes at the same time is calculated, the time-varying fluctuation characteristics of the difference data are extracted and recombined according to the sampling time order to generate an instantaneous difference sequence. S103: For the instantaneous difference sequence, calculate the gradient of numerical change at adjacent time points point by point, accumulate and calculate the absolute value of the gradient and calculate the average value, obtain the amplitude fluctuation rate index, integrate the tissue oxygen saturation time series data, instantaneous difference sequence and amplitude fluctuation rate index, and generate a microcirculation blood flow feature set.

4. The method for real-time monitoring of burn wound healing process according to claim 3, characterized in that, The specific steps for obtaining the global deformation vector field are as follows: S201: Acquire wound image data from continuous monitoring cycles, convert each frame to HSV color space and extract saturation component layers, set a preset threshold range matching the spectral characteristics of normal skin, perform binarization segmentation on the saturation component layers to obtain a binary mask image, perform morphological closing operation on the binary mask image, fill internal holes and smooth edge contours, extract the largest connected component as an effective mask, and generate an outer reference region. S202: For the outer reference area, filter the pixel coordinates of the corner response function values ​​that are greater than the preset feature threshold to obtain texture feature points, construct a multi-layer resolution image pyramid structure, search the matching coordinates with the smallest difference in local neighborhood gray level distribution between the texture feature points and the image at adjacent time points layer by layer from top to bottom, calculate the spatial vector difference between the matching coordinates and the original position, remove outlier vectors whose modulus exceeds the preset limit and aggregate the remaining results to generate displacement vector data. S203: Construct an irregular triangular mesh based on the coordinates of discrete feature points in the displacement vector data, establish a regular two-dimensional covering mesh based on the image pixel coordinate system, use radial basis function interpolation, calculate the estimated displacement of each node position in the two-dimensional covering mesh by weighting the displacement values ​​of the vertices of the irregular triangular mesh, map the estimated displacement to the corresponding mesh node coordinate index to establish a vector matrix, and generate a global deformation vector field.

5. The method for real-time monitoring of burn wound healing process according to claim 4, characterized in that, The process of setting a preset threshold range that matches the spectral characteristics of normal skin is as follows: Collect healthy skin sample images under the same lighting conditions as the wound image data, extract the pixel values ​​of the saturation channel in HSV space, calculate the expected value and standard deviation of all saturation channel pixel values, set the sum of the expected value and 3 times the standard deviation as the upper limit of the threshold, and set the difference between the expected value and 3 times the standard deviation as the lower limit of the threshold to generate the preset threshold range. The process of setting the preset feature threshold is as follows: Calculate the horizontal and vertical gradient components of the pixels within the outer reference region, construct the second-order moment matrix and calculate the corresponding corner response function values, count all corner response function values, calculate the segmentation value when the inter-class variance is maximized using the maximum inter-class variance method, and set it as the preset feature threshold.

6. The method for real-time monitoring of burn wound healing process according to claim 4, characterized in that, The specific steps for obtaining the regeneration shrinkage density coefficient are as follows: S301: Calculate the derivatives of pixel grayscale in the horizontal and vertical directions for wound image data, construct a texture gradient matrix and locate grayscale abrupt change regions, perform curve fitting on the pixel set to generate physical edge lines, track the coordinate changes of discrete sampling points on the physical edge lines in continuous time series, calculate the spatial vector difference between the current coordinates and the initial coordinates, and generate the total physical displacement vector. S302: Read the coordinates of the discrete sampling points associated with the total physical displacement vector, project the coordinates onto the spatial grid system of the global deformation vector field, calculate the background reference displacement vector at each projection point based on the grid node values, subtract the corresponding background reference displacement vector from the total physical displacement vector point by point, remove the background deformation interference component, and generate the net proliferation migration vector field. S303: Calculate the magnitude of each vector element in the net proliferation migration vector field to obtain the net proliferation magnitude sequence, simultaneously calculate the magnitude of the corresponding vector element in the total physical displacement vector to obtain the total displacement magnitude sequence, perform point-to-point division to calculate the ratio of the net proliferation magnitude to the total displacement magnitude, calculate the arithmetic mean of all the ratio values, and generate the regeneration shrinkage ratio coefficient.

7. The method for real-time monitoring of burn wound healing process according to claim 6, characterized in that, The process of locating the gray-scale abrupt change region is as follows: Extract the horizontal and vertical derivatives from the texture gradient matrix, calculate the square root of their sum to obtain the gradient magnitude, and calculate the arctangent of their ratio to obtain the gradient direction. Compare the gradient magnitude of the current pixel with that of its two adjacent pixels along the gradient direction. If the current pixel is not a local maximum, set the gradient magnitude to zero and perform non-maximum suppression. Statistically analyze the gradient magnitude distribution data after processing, calculate the segmentation value when the inter-class variance is maximized using the maximum inter-class variance method, set it as the adaptive edge segmentation threshold, and filter pixels with gradient magnitudes greater than the adaptive edge segmentation threshold to establish the gray-scale abrupt change region.

8. The method for real-time monitoring of burn wound healing process according to claim 6, characterized in that, The specific steps for obtaining the recursive feature matrix are as follows: S401: Call the microcirculation blood flow feature set, extract tissue oxygen saturation time series data, set the embedding dimension and time delay parameters, perform time series reconstruction processing on the tissue oxygen saturation time series data according to the time delay parameters, convert the one-dimensional time series data into discrete vectors in a multi-dimensional spatial coordinate system, arrange all discrete vectors according to the original sampling time order, and generate a state vector sequence. S402: Perform a full permutation traversal operation on the state vector sequence, extract the vector coordinates corresponding to any two time points, calculate the sum of squares of the differences between the two vectors in each dimension of the multidimensional space and perform a square root operation, obtain the Euclidean distance value of the vector pair at the corresponding time point, construct a two-dimensional square matrix with the time point as the row and column index, fill the Euclidean distance value into the corresponding coordinate position of the two-dimensional square matrix, and generate a vector spacing matrix. S403: Obtain the preset cutoff radius standard, traverse the Euclidean distance values ​​stored in the vector spacing matrix, compare the Euclidean distance values ​​with the preset cutoff radius standard item by item, filter the coordinate positions where the Euclidean distance value is less than the cutoff radius standard and perform binarization marking, and generate a recursive feature matrix.

9. The method for real-time monitoring of burn wound healing process according to claim 8, characterized in that, The specific steps for obtaining the wound healing assessment results are as follows: S501: Scan the recursive feature matrix to identify vertical line segment structures parallel to the main diagonal, calculate the numerical ratio of the total number of recursive points in the vertical line segment structure to the total number of recursive points in the whole matrix to obtain the laminar flow index, calculate the arithmetic mean of the lengths of all vertical line segment structures to obtain the capture time index, and generate a set of nonlinear dynamic parameters. S502: The amplitude fluctuation rate index is extracted by calling the microcirculation blood flow feature set, the regeneration contraction ratio coefficient is read, and multi-dimensional feature splicing is performed with the nonlinear dynamic parameter set. The spatial distance between the spliced ​​feature vector and the preset pathological benchmark vector is calculated. The current category label is determined based on the minimum distance, the pathological state classification identifier is obtained, and the wound healing assessment result is generated.

10. A real-time monitoring system for the healing process of burn wounds, characterized in that, The system is used to implement the real-time monitoring method for burn wound healing process according to any one of claims 1-9, the system comprising: Signal feature extraction module: Collects oxygen saturation data and locates extreme points, interpolates and connects extreme points to construct an envelope, calculates the envelope difference to obtain the instantaneous difference sequence, calculates the instantaneous difference gradient to obtain the amplitude fluctuation rate index, and constructs a microcirculation blood flow feature set; Deformation field dynamic tracking module: By segmenting the wound image by pixel color features, dividing the outer reference area and filtering the texture feature points within the area, calculating the displacement vector between multiple time points and interpolating it, a global deformation vector field is constructed. Regeneration weight analysis module: Fit the texture gradient of the wound image to the edge line, calculate the total displacement vector of the edge line, and call the global deformation vector field to perform vector subtraction to obtain the net proliferation migration vector field, calculate the ratio with the total physical displacement vector magnitude, and generate the regeneration shrinkage weight coefficient; The time-series recursive analysis module calls the microcirculation blood flow feature set, extracts tissue oxygen saturation time-series data, constructs a state vector sequence, calculates the Euclidean distance between each pair of time vectors in the state vector sequence, filters out coordinate positions with Euclidean distances less than the preset cutoff radius standard and performs binarization marking, and generates a recursive feature matrix. Healing process determination module: Scans the vertical segments of the recursive feature matrix, calculates the proportion of vertical segment points to obtain laminar flow index and average length to obtain capture time index, combines the amplitude fluctuation rate index in the microcirculation blood flow feature set and the regeneration contraction ratio coefficient to make a determination, and outputs the wound healing assessment result.

Images