Pressure damage detection method based on fractional order composite multi-scale sample entropy

Abstract

The invention belongs to the field of medical image analysis, and relates to a pressure damage detection method based on fractional order composite multi-scale sample entropy, which comprises the following steps: dynamically tracking a region of interest of a pressure damage video based on an adaptive optical flow field to obtain a tracked video of the region of interest; processing the tracking video through an Euler video amplification method, and constructing a brightness image sequence; performing fractional order differential processing on the brightness image sequence, calculating a multi-scale sample entropy feature value, and constructing a sample entropy feature vector based on the multi-scale sample entropy feature value; based on the sample entropy feature vector, constructing a training sample set to train a support vector machine classifier, and obtaining a stress damage staging model; and inputting a to-be-detected pressure damage video into the pressure damage staging model to obtain a damage staging result of the to-be-detected pressure damage video so as to improve the precision and efficiency of pressure damage detection.

Description

Technical Field

[0001] The present invention relates to the field of medical image analysis, and specifically discloses a pressure injury detection method based on fractional composite multi-scale sample entropy. Background Art

[0002] Pressure injury, also known as pressure ulcer, is an injury caused by long-term local tissue compression, resulting in blood circulation disorders, and then ischemia, hypoxia, and malnutrition of the skin and subcutaneous tissues. It is common in patients who are bedridden or in wheelchairs for a long time, and often occurs in bony prominences such as the sacrococcygeal region and heels. Early identification and staging of pressure injuries are crucial for preventing their further deterioration. However, existing detection methods still have many limitations.

[0003] Current clinical detection mainly relies on static image recognition techniques. For example, the erythema threshold segmentation based on the HSV color model can identify the erythema area and combine morphological processing to locate the suspected pressure injury position; or the gray-level co-occurrence matrix (GLCM) is used to quantify the roughness and uniformity of skin texture to distinguish normal tissues from damaged areas. In addition, near-infrared spectroscopy or thermal imaging techniques are also used to judge local ischemia through temperature differences. However, these methods are all based on single-frame image analysis, which can only capture instantaneous static features and cannot reflect the dynamic evolution process of pressure injuries. For example, the early erythema of stage I pressure injury may show periodic brightness changes with the fluctuation of microcirculation. Static images are likely to misjudge it as transient congestion, and traditional image entropy methods are sensitive to lighting conditions, resulting in a significant decrease in specificity in patients with large skin color differences. Although existing studies have tried to introduce time series analysis methods such as multi-scale entropy, their fixed parameter and mean coarse-graining strategies can neither effectively retain the high-frequency details of the signal (such as the transient fluctuations in the erythema area, being insensitive to weak dynamic changes), nor can they adaptively adjust parameters to adapt to individual skin characteristic differences, resulting in limited judgment accuracy for pressure injury staging. Therefore, there is an urgent need for a method that integrates dynamic signal analysis to make up for the deficiencies of static image technology and achieve accurate staging and real-time monitoring of pressure injuries.

[0004] In view of this, the present invention provides a pressure injury detection method based on fractional composite multi-scale entropy (FrCMSE) to improve the accuracy and efficiency of pressure injury detection. Summary of the Invention

[0005] The object of the present invention is to provide a pressure injury detection method based on fractional-order composite multi-scale sample entropy. The specific scheme includes: dynamically tracking the region of interest of the pressure injury video based on the adaptive optical flow field to obtain the tracking video of the region of interest; processing the tracking video by the Euler video magnification method to construct a luminance image sequence; performing fractional-order differential processing on the luminance image sequence, calculating multi-scale sample entropy eigenvalues, and constructing a sample entropy feature vector based on the multi-scale sample entropy eigenvalues; constructing a training sample set to train a support vector machine classifier based on the sample entropy feature vector to obtain a pressure injury staging model; inputting the pressure injury video to be measured into the pressure injury staging model to obtain the injury staging result of the pressure injury video to be measured.

[0006] Further, the dynamically tracking the region of interest of the pressure injury video based on the adaptive optical flow field includes: determining the image luminance variance based on the degree of illumination change of the video frame image; determining the image gradient mean based on the degree of motion complexity of the video frame image; determining the adaptive weight coefficient of the adaptive optical flow field based on the image luminance variance and the image gradient mean; constructing an optimization objective function of the luminance consistency error of the adaptive optical flow field based on the adaptive weight coefficient.

[0007] Further, the calculation formula of the adaptive weight coefficient is: ; The calculation formula of the optimization objective function of the luminance consistency error is: ; Wherein, represents the adaptive weight coefficient; and respectively represent the luminance weight coefficient and the gradient weight coefficient; represents the image luminance variance of the frame image in the pressure injury video; represents the image gradient mean of the frame image in the pressure injury video; represents the luminance value of the pressure injury video; represents the spatial gradient vector of the luminance; represents the gradient norm; represents the value of the optimization objective function of the luminance consistency error; i represents the pixel point variable in the region of interest; N represents the total number of pixel points in the region of interest; represents the luminance gradient of the frame image in the x direction in the pressure injury video; represents the luminance gradient of the frame image in the y direction in the pressure injury video; represents the temporal gradient of the luminance of the frame image in the pressure injury video; represents the displacement of the i-th pixel point variable in the x direction; denotes the displacement of the \(i\)-th pixel variable in the \(y\) direction; denotes the displacement gradient of the \(i\)-th pixel variable in the \(x\) direction; denotes the displacement gradient of the \(i\)-th pixel variable in the \(y\) direction.

[0008] Further, the fractional-order differential processing of the luminance image sequence, calculating the multi-scale sample entropy eigenvalues, and constructing a sample entropy feature vector based on the multi-scale sample entropy eigenvalues includes: nonlinearly enhancing the luminance image sequence through fractional-order differential to obtain fractional-order derivatives at multiple time points, arranging the fractional-order derivatives to obtain a fractional-order derivative sequence; respectively performing coarse-graining processing on the fractional-order derivative sequence based on multiple scale factors to obtain multi-scale coarse-grained sequences; calculating the multi-scale sample entropy of the multi-scale coarse-grained sequences to obtain the sample entropy feature vector.

[0009] Further, the formula for nonlinearly enhancing the luminance image sequence through fractional-order differential to obtain fractional-order derivatives at multiple time points is: ; where denotes the fractional-order derivative at the time point ; \(t\) represents the time variable; denotes the luminance image sequence at the time point \(t\); denotes the value of the calculation formula when the sampling time interval approaches 0 from a direction greater than 0; \(h\) represents the sampling time interval; denotes the fractional-order; \(k\) represents the backtracking time length; denotes the lower limit parameter; denotes the generalized binomial coefficient; denotes the time point ;

[0010] Further, the process of respectively performing coarse-graining processing on the fractional-order derivative sequence based on multiple scale factors to obtain a coarse-grained sequence includes: determining multiple scale factors, and respectively dividing the fractional-order derivative sequence into multiple non-overlapping time windows of multiple time scales according to the time lengths of the multiple scale factors; respectively calculating the average value of the fractional-order derivatives within each non-overlapping time window to obtain coarse-grained data points; respectively arranging the coarse-grained data points for each scale factor to obtain a coarse-grained sequence; arranging the coarse-grained sequences of multiple scales to obtain a multi-scale coarse-grained sequence.

[0011] Further, performing multi-scale sample entropy calculation on the multi-scale coarse-grained sequence to obtain the sample entropy feature vector includes: calculating the sample entropy of each coarse-grained sequence in the multi-scale coarse-grained sequence respectively, including: constructing an embedding vector of the first dimension to obtain a first number of first templates; calculating the distances between the first templates, and taking the number of first templates with distances less than the distance tolerance as the first numerical parameter; constructing an embedding vector of the second dimension to obtain a second number of second templates; calculating the distances between the second templates, and taking the number of second templates with distances less than the distance tolerance as the second numerical parameter; calculating the sample entropy based on the first numerical parameter and the second numerical parameter; calculating the composite sample entropy under multiple scale factors respectively based on the sample entropy; arranging the composite sample entropy to obtain the sample entropy feature vector.

[0012] Further, it also includes performing weighted processing on the composite sample entropy under multiple scale factors in the sample entropy feature vector to obtain a weighted sample entropy feature vector, including: dividing the value range of the composite sample entropy to obtain multiple pressure injury categories; respectively determining the target pressure injury categories into which multiple composite sample entropies fall based on the values of the composite sample entropy; calculating the correlation between the composite sample entropy and multiple pressure injury categories based on the pressure injury category and the target pressure injury category; respectively determining the redundancy between the composite sample entropies based on the target pressure injury category; calculating the weight score based on the correlation and its corresponding redundancy; normalizing the weight score to obtain the weighted weight of each composite sample entropy; taking the product of the weighted weight and the corresponding composite sample entropy as the weighted sample entropy, and arranging the weighted sample entropy to obtain the weighted sample entropy feature vector.

[0013] Further, the calculation formula for the correlation is: ; where represents the correlation between the first composite sample entropy and multiple pressure injury categories respectively; represents the target pressure injury category belonging to the target pressure injury category to which the first composite sample entropy belongs; represents the composite sample entropy of the th scale; represents the first composite sample entropy variable; represents that the pressure injury category c belongs to multiple pressure injury categories ; represents the set of pressure injury categories, including multiple types of pressure injury categories; Denote the joint probability that the target pressure injury category of the first composite sample entropy is a and the pressure injury category is c; Denote the first composite sample entropy The target pressure injury category of Marginal probability; Denote the marginal probability that the pressure injury category is c; The calculation formula of the redundancy is: ; Where Denote the redundancy of the first composite sample entropy ; Denote the second composite sample entropy variable; Denote that the target pressure injury category b belongs to the target pressure injury category to which the second composite sample entropy belongs; Denote the first composite sample entropy The target pressure injury category of And the second composite sample entropy The joint probability that the target pressure injury category is b; Denote the second composite sample entropy The marginal probability that the target pressure injury category is b; The calculation formula of the weight score is: .

[0014] Furthermore, constructing a training sample set based on the sample entropy feature vector to train a support vector machine classifier to obtain a pressure injury staging model, including: Construct the training sample set based on the sample entropy feature vector; Take the radial basis function as the kernel function of the support vector machine classifier; the expression of the kernel function is: ; Where Denote the kernel function; Denote the th weighted sample entropy feature vector; Denote the th weighted sample entropy feature vector; exp represents the exponential function; Denote the kernel function parameter; ‖*‖ represents the Euclidean norm; When training the support vector machine classifier, find the optimal hyperplane by optimizing the model objective function to obtain the pressure injury staging model; the expression of the model objective function is: ; In the formula, Denote the minimization of the hyperplane normal vector \(W\) and the bias term \(d\); \(W\) represents the hyperplane normal vector; \(d\) represents the bias term; Denote the regularization parameter; \(q\) represents the training sample variable; \(Q\) represents the total number of training samples; Denote the slack variable.

[0015] The present invention preprocesses the input optical flow tracking ROI video data through Euler video magnification to enhance the patient's skin micro-vibration signal and capture the minute changes in the luminance signal.

[0016] The present invention performs fractional-order differential processing on the magnified ROI image sequence, calculates its FrCMSE value, and extracts the FrCMSE feature vector reflecting the dynamic change characteristics of pressure injury. Fractional-order differential is an extension of traditional integer-order calculus, allowing the differential order to be any real number, and can better describe signals with non-linear and non-stationary characteristics. In pressure injury video detection, fractional-order differential non-linearly enhances the ROI (region of interest) image sequence, highlights high-frequency details (such as skin texture, minute changes in early pressure injury), and non-linearly retains low-frequency information (such as the gray-scale gradient of smooth skin regions). This processing can effectively capture the dynamic change characteristics of pressure injury while suppressing noise interference.

[0017] The present invention calculates the weight of each FrCMSE feature through the minimum redundancy (mRMR) algorithm, which can maximize the correlation between the feature and the target category and improve the recognition accuracy. To facilitate the calculation of the mutual information between the feature and the class label, the present invention performs binning (discretization) processing on the extracted FrCMSE feature vector sample set. Considering that the FrCMSE feature is a continuous-valued variable, directly participating in the mutual information calculation will lead to unstable probability estimation, especially more likely to introduce bias when the sample size is limited. Therefore, a discretization method based on interval partitioning is adopted to map the continuous feature values to a finite number of discrete states, thus facilitating the estimation of probability distribution and subsequent statistical operations.

[0018] The present invention uses the processed WFrCMSE feature vector sample set to train an SVM (support vector machine) classifier to construct a pressure injury staging model. SVM realizes classification by finding the optimal hyperplane, maximizing the margin between categories, thereby improving the generalization ability of the classifier. In pressure injury detection, the radial basis function (RBF) kernel is selected as the kernel function of SVM. The RBF kernel can effectively handle non-linear classification problems, especially suitable for data classification in high-dimensional feature spaces. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 FIG. is an exemplary flowchart of a pressure injury detection method based on fractional-order composite multi-scale sample entropy proposed by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention usually described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0021] Figure 1 It is an exemplary flowchart of a pressure injury detection method based on fractional-order composite multi-scale sample entropy proposed by the present invention. As Figure 1 shown, the pressure injury detection method based on fractional-order composite multi-scale sample entropy includes the following: Dynamically track the region of interest of the pressure injury video based on the adaptive optical flow field to obtain the tracking video of the region of interest. In practical applications, the brightness consistency constraint may fail due to factors such as illumination changes and noise interference. To improve the robustness of the optical flow field calculation, through the adaptive optical flow field optimization algorithm, the parameters of the optical flow calculation are dynamically adjusted according to the motion complexity and illumination conditions of the video frames. The region of interest may refer to the region in the video where pressure injury may exist. The tracking video refers to the video obtained by tracking the region of interest.

[0022] In some embodiments, the adaptive optical flow field optimization algorithm dynamically adjusts the weight coefficients of the optical flow field calculation by analyzing the image brightness variance and gradient information to better adapt to the motion tracking requirements, including: determining the image brightness variance based on the degree of illumination change of the video frame image. Determining the average image gradient based on the motion complexity of the video frame image. The video frame image refers to the frame image in the pressure injury video. Determining the adaptive weight coefficient of the adaptive optical flow field based on the image brightness variance and the average image gradient. In some embodiments, the calculation formula of the adaptive weight coefficient is: ; where represents the adaptive weight coefficient, which is used to adjust the balance between the smooth term and the data term of the optical flow field calculation; and respectively represent the brightness weight coefficient and the gradient weight coefficient, which are used to balance the influence of the brightness variance and the average gradient on the optical flow field calculation; represents the image brightness variance of the frame image in the pressure injury video, which reflects the degree of change in the image brightness. The larger the brightness variance, the more obvious the illumination change in the image, and the stronger the smoothness required for the optical flow field calculation; Represents the mean image gradient of the frame images in the pressure injury video, reflecting the intensity of edges and textures in the image. The larger the gradient mean, the higher the motion complexity in the image, and higher precision is required for optical flow field calculation; Represents the luminance value of the pressure injury video; Represents the spatial gradient vector of luminance; Represents the gradient magnitude.

[0023] Based on the adaptive weight coefficient, an optimization objective function for the luminance consistency error of the adaptive optical flow field is constructed. The calculation formula of the optimization objective function for the luminance consistency error is: ; Where Represents the value of the optimization objective function for the luminance consistency error; i represents the pixel point variable in the region of interest; N represents the total number of pixel points in the region of interest; Represents the luminance gradient of the frame image in the x direction in the pressure injury video; Represents the luminance gradient of the frame image in the y direction in the pressure injury video; Represents the temporal gradient of the frame image luminance in the pressure injury video; Represents the displacement of the i-th pixel point variable in the x direction; Represents the displacement of the i-th pixel point variable in the y direction; Represents the displacement gradient of the i-th pixel point variable in the x direction; Represents the displacement gradient of the i-th pixel point variable in the y direction. and are smoothing terms Process the tracking video through the Euler video magnification method to construct a luminance image sequence. The luminance image sequence refers to an image sequence related to the luminance of the image obtained after processing by the Euler method.

[0024] Perform fractional-order differential processing on the luminance image sequence, calculate the multi-scale sample entropy eigenvalues, and construct a sample entropy feature vector based on the multi-scale sample entropy eigenvalues. In some embodiments, performing fractional-order differential processing on the luminance image sequence, calculating the multi-scale sample entropy eigenvalues, and constructing a sample entropy feature vector based on the multi-scale sample entropy eigenvalues includes: performing non-linear enhancement on the luminance image sequence through fractional-order differentiation to obtain the fractional-order derivatives at multiple time points, and arranging the fractional-order derivatives to obtain a fractional-order derivative sequence. In some embodiments, the calculation formula for performing non-linear enhancement on the luminance image sequence through fractional-order differentiation to obtain the fractional-order derivatives at multiple time points is: ; Where Represents at the time point The fractional derivative at; t represents the time variable; represents the sequence of luminance images at time point t; represents the value of the calculation formula when the sampling time interval approaches 0 from a direction greater than 0; h represents the sampling time interval, which takes ; represents the fractional order, controlling the signal strength; k represents the length of the backtracking time, and its value range is In this method, k = 30 is taken, and K represents the total length of the signal sequence; represents the lower limit parameter; represents the time point of the sequence of luminance images; represents the generalized binomial coefficient, which is calculated through the recurrence formula: .

[0025] Coarsen the fractional derivative sequence respectively based on multiple scale factors to obtain a multi-scale coarsened sequence. The scale factor ( ) is defined as: represents the granularity of analysis. For example, when = 3, the original time series is divided into multiple non-overlapping windows with a length of 3. Coarsened sequence generation: Calculate the average value of the data within each window to generate a new coarsened sequence. For example, the original sequence 2, 4, 6, 8, 10, 122, 4, 6, 8, 10, 12 becomes 3, 7, 113, 7, 11 when = 2, that is, take the average of every two data points.

[0026] In some embodiments, the coarsening the fractional derivative sequence respectively based on multiple scale factors to obtain a coarsened sequence includes: determining multiple scale factors, and dividing the fractional derivative sequence into multiple non-overlapping time windows of multiple time scales according to the time lengths of the multiple scale factors. For example, the fractional derivative sequence can be decomposed into multiple time windows at a fixed time interval. Calculate the average value of the fractional derivatives within each non-overlapping time window respectively to obtain coarsened data points. Arrange the coarsened data points for each scale factor respectively to obtain a coarsened sequence. For example, for the time series , the coarsened data points under the scale factor are defined as: ; where j is the sequence number of the coarsened sequence, and i represents the order of the original time series before coarsening; T is the length of the original sequence; represents the i-th data point of the original time series; is the scale factor, and the final coarse-grained sequence can be expressed as .

[0027] Arrange the coarse-grained sequences at multiple scales to obtain a multi-scale coarse-grained sequence. Calculate the multi-scale sample entropy of the multi-scale coarse-grained sequence to obtain the sample entropy feature vector, including: calculating the sample entropy of each coarse-grained sequence in the multi-scale coarse-grained sequence respectively, including: constructing an embedding vector of the first dimension to obtain a first number of first templates. For example, construct an embedding vector of dimension m: ; There are a total of N - m templates.

[0028] Calculate the distance between the first templates, and take the number of first templates with a distance less than the distance tolerance as the first parameter. For example, calculate the distance between templates and count the number of those less than the tolerance r (usually r = 0.15×standard deviation) and denote it as B.

[0029] ; Count the number of, and denote it as B.

[0030] Construct an embedding vector of the second dimension to obtain a second number of second templates. Calculate the distance between the second templates, and take the number of second templates with a distance less than the distance tolerance as the second parameter. For example, when the dimension of the embedding vector becomes m + 1, repeat the above steps for determining the number of templates and denote the number as A.

[0031] Based on the first parameter and the second parameter, calculate the sample entropy; the formula for calculating the sample entropy based on the first parameter and the second parameter is: ; where represents the sample entropy when the embedding vector is of dimension m and the distance tolerance is r; A represents the second parameter; B represents the first parameter. If the sequence remains similar after the dimension of the embedding vector increases , it indicates that the sequence is stable and the entropy value is low. If the number of matches drops significantly after the dimension of the embedding vector increases , it indicates that the sequence is very unstable, the entropy value is high, and the complexity is high.

[0032] Based on the sample entropy, calculate the composite sample entropy under multiple scale factors respectively; the formula for the composite sample entropy is: ; where represents the composite sample entropy of the scale factor; Represents the scale factor variable; Represents the maximum value of the scale factor variable; Represents the coarse-grained sequence of sample entropy; Represents when the scale factor is the coarse-grained sequence.

[0033] Arrange the composite sample entropy to obtain the sample entropy feature vector. Calculation process of FrCMSE: First, perform fractional-order differential processing on the luminance signal, then perform coarse-graining processing, and calculate the composite sample entropy at different scales to finally construct the FrCMSE feature vector from the sample entropies calculated at consecutive scales. The calculation formula of the sample entropy feature vector is: ; ; wherein, represents the sample entropy feature vector; , and respectively represent the composite sample entropies of the first scale factor, the second scale factor, and the scale factor.

[0034] In some embodiments, it further includes weighting the composite sample entropies at multiple scale factors in the sample entropy feature vector to obtain a weighted sample entropy feature vector, including: Dividing the value range of the composite sample entropy to obtain multiple pressure injury categories; based on the value of the composite sample entropy, respectively determine the target pressure injury categories into which multiple composite sample entropies fall. For each feature dimension , first count the minimum value and the maximum value that appear in the entire sample set, and calculate the feature value range. Then evenly divide this range into V sub-intervals (usually taking V = 5 - 10), and each sub-interval corresponds to a discrete value number . For the feature value of a certain dimension of any sample, determine which interval it belongs to and map it to the corresponding discrete number.

[0035] Based on the pressure injury category and the target pressure injury category, calculate the correlation between the composite sample entropy and multiple pressure injury categories; calculate the correlation between the pressure injury classification label and the single-dimensional data of the feature vector. The calculation formula of the correlation is: ; wherein, represents the correlations between the first composite sample entropy and multiple pressure injury categories respectively; Indicates the target pressure injury category Belongs to the target pressure injury category to which the first composite sample entropy belongs; Indicates the Composite sample entropy of the scale; Indicates the first composite sample entropy variable; Indicates that the pressure injury category c belongs to multiple pressure injury categories In this task of pressure injury detection, there are a total of 5 categories (healthy, stage 1 pressure injury, stage 2 pressure injury, stage 3 pressure injury, stage 4 pressure injury); Indicates the set of pressure injury categories, including multiple pressure injury categories; Indicates the joint probability that the target pressure injury category of the first composite sample entropy is a and the pressure injury category is c; Indicates the first composite sample entropy The target pressure injury category of which is Marginal probability; Indicates the marginal probability that the pressure injury category is c.

[0036] Based on the target pressure injury category, determine the redundancy between the composite sample entropies respectively; eliminate the duplicate information between features to ensure strong independence of features within the subset. The calculation formula for the redundancy is: ; Wherein, Indicates the redundancy of the first composite sample entropy ; Indicates the second composite sample entropy variable; Indicates that the target pressure injury category b belongs to the target pressure injury category to which the second composite sample entropy belongs; Indicates the first composite sample entropy The target pressure injury category of which is And the second composite sample entropy The joint probability that the target pressure injury category is b; Indicates the marginal probability that the target pressure injury category of the second composite sample entropy Is b.

[0037] Based on the correlation and its corresponding redundancy, calculate the weight score; the calculation formula for the weight score is: .

[0038] Normalize the weight score to obtain the weighted weight of each composite sample entropy.

[0039] Take the product of the weighted weight and the corresponding composite sample entropy as the weighted sample entropy, and arrange the weighted sample entropy to obtain a weighted sample entropy feature vector. Finally, convert the FrCMSE feature vector into a weighted feature vector WFrCMSE: ; Construct a training sample set based on the sample entropy feature vector to train a support vector machine classifier, and obtain a pressure injury staging model. The constructing a training sample set based on the sample entropy feature vector to train a support vector machine classifier and obtaining a pressure injury staging model includes: Construct the training sample set based on the sample entropy feature vector.

[0040] Take the radial basis function as the kernel function of the support vector machine classifier; the expression of the kernel function is: ; where, represents the kernel function; represents the th weighted sample entropy feature vector; represents the th weighted sample entropy feature vector; exp represents the exponential function; represents the kernel function parameter, which is used to control the width of the kernel function; ‖*‖ represents the Euclidean norm.

[0041] When training the support vector machine classifier, find the optimal hyperplane by optimizing the model objective function to obtain the pressure injury staging model; the expression of the model objective function is: ; In the formula, represents minimizing the hyperplane normal vector W and the bias term d; W represents the hyperplane normal vector; d represents the bias term; represents the regularization parameter, which controls the fault tolerance ability of the classifier. A larger value will reduce the training error, but may lead to overfitting; q represents the training sample variable; Q represents the total number of training samples; represents the slack variable, which is used to handle classification errors.

[0042] Input the pressure injury video to be measured into the pressure injury staging model to obtain the injury staging result of the pressure injury video to be measured. For new video data, use the trained pressure injury detection model for automatic detection and output the detection result.

[0043] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A pressure injury detection method based on fractional-order composite multi-scale sample entropy, characterized in that Including: Dynamically tracking the region of interest of the pressure injury video based on the adaptive optical flow field to obtain the tracking video of the region of interest; Processing the tracking video by the Euler video magnification method to construct a luminance image sequence; Performing fractional-order differential processing on the luminance image sequence, calculating multi-scale sample entropy eigenvalues, and constructing a sample entropy feature vector based on the multi-scale sample entropy eigenvalues; Constructing a training sample set to train a support vector machine classifier based on the sample entropy feature vector to obtain a pressure injury staging model; Inputting the pressure injury video to be measured into the pressure injury staging model to obtain the injury staging result of the pressure injury video to be measured.

2. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 1, characterized in that The dynamically tracking the region of interest of the pressure injury video based on the adaptive optical flow field includes: Determining the image luminance variance based on the degree of illumination change of the video frame image; Determining the image gradient mean based on the degree of motion complexity of the video frame image; Determining the adaptive weight coefficient of the adaptive optical flow field based on the image luminance variance and the image gradient mean; Constructing an optimization objective function for the luminance consistency error of the adaptive optical flow field based on the adaptive weight coefficient.

3. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 2, wherein The calculation formula of the adaptive weight coefficient is: ； The calculation formula of the optimization objective function for the luminance consistency error is: ； Among them, represents the adaptive weight coefficient; and represent the luminance weight coefficient and the gradient weight coefficient respectively; represents the image luminance variance of the frame image in the pressure damage video; represents the image gradient mean of the frame image in the pressure damage video; represents the luminance value of the pressure damage video; represents the spatial gradient vector of luminance; represents the gradient magnitude; represents the value of the optimization objective function of the luminance consistency error; i represents the pixel variable within the region of interest; N represents the total number of pixels within the region of interest; represents the luminance gradient of the frame image in the x direction in the pressure damage video; represents the luminance gradient of the frame image in the y direction in the pressure damage video; represents the temporal gradient of the frame image luminance in the pressure damage video; represents the displacement of the i-th pixel variable in the x direction; represents the displacement of the i-th pixel variable in the y direction; represents the displacement gradient of the i-th pixel variable in the x direction; represents the displacement gradient of the i-th pixel variable in the y direction.

4. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 1, wherein The performing fractional-order differential processing on the luminance image sequence, calculating multi-scale sample entropy eigenvalues, and constructing a sample entropy feature vector based on the multi-scale sample entropy eigenvalues includes: Performing non-linear enhancement on the luminance image sequence through fractional-order differential to obtain fractional-order derivatives at multiple time points, and arranging the fractional-order derivatives to obtain a fractional-order derivative sequence; Performing coarse-graining processing on the fractional-order derivative sequence respectively based on multiple scale factors to obtain multi-scale coarse-grained sequences; Performing multi-scale sample entropy calculation on the multi-scale coarse-grained sequences to obtain the sample entropy feature vector.

5. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 4, characterized in that The calculation formula for performing non-linear enhancement on the luminance image sequence through fractional-order differential to obtain fractional-order derivatives at multiple time points is: ； Among them, represents the fractional derivative at the time point ; t represents the time variable; represents the sequence of luminance images at the time point t; represents the value of the calculation formula when the sampling time interval approaches 0 from a direction greater than 0; h represents the sampling time interval; represents the fractional order; k represents the length of the backtracking time; represents the lower limit parameter; represents the generalized binomial coefficient; represents the time point at which the sequence of luminance images is located.

6. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 4, wherein The performing coarse-graining processing on the fractional-order derivative sequence respectively based on multiple scale factors to obtain coarse-grained sequences includes: Determining multiple scale factors, and dividing the fractional-order derivative sequence into multiple non-overlapping time windows of multiple time scales according to the time lengths of the multiple scale factors respectively; Calculating the average value of the fractional-order derivatives in each non-overlapping time window respectively to obtain coarse-grained data points; Arranging the coarse-grained data points of each scale factor respectively to obtain coarse-grained sequences; Arranging the coarse-grained sequences of multiple scales to obtain multi-scale coarse-grained sequences.

7. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 6, wherein The performing multi-scale sample entropy calculation on the multi-scale coarse-grained sequences to obtain the sample entropy feature vector includes: Calculating the sample entropy of each coarse-grained sequence in the multi-scale coarse-grained sequences respectively, including: Constructing an embedding vector of the first dimension to obtain a first number of first templates; Calculating the distance between the first templates, and taking the number of first templates with a distance less than the distance tolerance as the first number parameter; Constructing an embedding vector of the second dimension to obtain a second number of second templates; Calculate the distance between the second templates, and use the number of second templates with a distance less than the distance tolerance as the second numerical parameter; Calculate the sample entropy based on the first numerical parameter and the second numerical parameter; Calculate the composite sample entropy at multiple scale factors based on the sample entropy; Arrange the composite sample entropy to obtain the sample entropy feature vector.

8. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 4, characterized in that It further includes performing a weighting process on the composite sample entropy at multiple scale factors in the sample entropy feature vector to obtain a weighted sample entropy feature vector, including: Divide the value range of the composite sample entropy to obtain multiple pressure injury categories; Based on the value of the composite sample entropy, respectively determine the target pressure injury categories into which multiple composite sample entropies fall; Calculate the correlation between the composite sample entropy and multiple pressure injury categories based on the pressure injury category and the target pressure injury category; Based on the target pressure injury category, respectively determine the redundancy between the composite sample entropies; Calculate the weight score based on the correlation and its corresponding redundancy; Normalize the weight score to obtain the weighted weight of each composite sample entropy; Use the product of the weighted weight and the corresponding composite sample entropy as the weighted sample entropy, and arrange the weighted sample entropy to obtain the weighted sample entropy feature vector.

9. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 8, characterized in that The calculation formula for the correlation is: ； Among them, represents the first composite sample entropy and its correlations with multiple pressure injury categories; represents the target pressure injury category which belongs to the target pressure injury category to which the first composite sample entropy belongs; represents the composite sample entropy at the scale; represents the first composite sample entropy variable; ; represents the set of pressure injury categories, including multiple pressure injury categories; represents the joint probability that the target pressure injury category of the first composite sample entropy is a and the pressure injury category is c; represents the first composite sample entropy whose target pressure injury category is the marginal probability; represents the marginal probability that the pressure injury category is c; The calculation formula for the redundancy is: ； Among them, represents the redundancy of the first composite sample entropy ; represents the second composite sample entropy variable; represents that the target pressure injury category b belongs to the target pressure injury category to which the second composite sample entropy belongs; represents the first composite sample entropy whose target pressure injury category is and the second composite sample entropy whose target pressure injury category is b; represents the marginal probability that the target pressure injury category of the second composite sample entropy is b; The calculation formula for the weight score is: 。 10. The pressure injury detection method based on fractional-order composite multi-scale sample entropy according to claim 1, wherein Constructing a training sample set based on the sample entropy feature vector to train a support vector machine classifier to obtain a pressure injury staging model, including: Construct the training sample set based on the sample entropy feature vector; Use the radial basis function as the kernel function of the support vector machine classifier; the expression of the kernel function is: ； Among them, represents a kernel function; represents the th weighted sample entropy feature vector; represents the th weighted sample entropy feature vector; exp represents the exponential function; represents the kernel function parameter; ‖*‖ represents the Euclidean norm; When training the support vector machine classifier, find the optimal hyperplane by optimizing the model objective function to obtain the pressure injury staging model; the expression of the model objective function is: ； In the formula, represents minimizing with respect to the hyperplane normal vector W and the bias term d; W represents the hyperplane normal vector; d represents the bias term; represents the regularization parameter; q represents the training sample variable; Q represents the total number of training samples; represents the slack variable.

Images