OCTA image-based nevus flammeus vascular structure analysis method and system

Abstract

The invention relates to an OCTA image-based nevus flammeus vascular structure analysis method and system. The method comprises the following steps: performing three-dimensional scanning on a nevus flammeus area before treatment based on an OCTA system to generate a tomography image and a projection image; fusing the vertical gradient component and the horizontal gradient component, and calculating a first epidermal layer thickness through a Dijkstra algorithm; carrying out adaptive threshold segmentation on the projection drawing, extracting a capillary network binary image, a blood vessel outer contour image and a skeleton image in combination with median filtering and a Canny operator, and calculating the average blood vessel diameter, the average blood vessel density, the length ratio of blood vessels with different tube diameters and other blood vessel structure features; and obtaining a second vascular structure feature and a second epidermal layer thickness after treatment, and performing curative effect analysis through a Logistic regression model in combination with the first epidermal layer thickness and the first vascular structure feature. According to the method, the accuracy and comprehensiveness of quantitative analysis of the nevus flammeus lesion structure are remarkably improved, and a scientific basis is provided for personalized formulation and curative effect prediction of a V-PDT treatment scheme.

Description

Technical Field

[0001] This invention belongs to the field of computer image processing technology, and in particular relates to a method and system for analyzing the vascular structure of port-wine stains based on OCTA images. Background Technology

[0002] Port-wine stains (PWS) are a congenital microvascular malformation of the skin, commonly seen in infants and young children, manifesting as pink, red, or purple patches on the skin surface. The main pathological feature of PWS is the abnormal dilation and malformation of capillaries in the superficial dermis. Vascular-targeted photodynamic therapy (V-PDT), as an emerging treatment method, has achieved significant efficacy in the treatment of PWS in recent years. However, current clinical assessment of PWS lesion structure mainly relies on the physician's visual judgment, lacking precise quantitative methods, thus limiting the personalized development of V-PDT treatment plans and the prediction of efficacy.

[0003] Furthermore, Optical Coherence Tomography Angiography (OCTA) is a novel non-invasive optical imaging technique that can provide high-resolution three-dimensional imaging of the skin's microvascular network. OCTA utilizes the principle of optical coherence to detect signal changes caused by blood cell movement, achieving non-contact, non-destructive imaging of the skin's microvascular structure. Compared to traditional histopathological biopsies, OCTA offers advantages such as being non-invasive, real-time, and rapid, providing more accurate vascular structural information for the diagnosis and treatment of peripheral skin diseases (PWS).

[0004] In recent years, OCTA-based image analysis technology has gradually gained attention in the diagnosis and treatment of peripheral blood vessel diseases (PWS). However, existing OCTA image analysis methods are often based on two-dimensional imaging, which limits the extraction of vascular depth information. In addition, existing quantitative indicators of vascular structure mainly focus on parameters such as vascular density and diameter, while the analysis of features such as the layered structure of blood vessels and the proportion of vertical vessels is still insufficient. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and system for analyzing the vascular structure of port-wine stains based on OCTA images to address the aforementioned technical issues, aiming to improve the accuracy and comprehensiveness of PWS lesion structure assessment.

[0006] In a first aspect, this application provides a method for analyzing the vascular structure of port-wine stains based on OCTA images, the method comprising:

[0007] The OCTA system based on the OMAG algorithm was used to perform a three-dimensional scan of the port-wine stain area before treatment, generating a tomographic scan and a projection image.

[0008] For the tomographic scan image, the vertical gradient component and the horizontal gradient component are fused, and the epidermal-air interface and the epidermal-dermal interface are located by Dijkstra's algorithm to calculate the thickness of the first epidermal layer.

[0009] Adaptive threshold segmentation is performed on the projection map, and median filtering and Canny operator are combined to extract the binary map of microvascular network, the outer contour map of blood vessels, and the skeleton map. Based on the binary map of microvascular network, the outer contour map of blood vessels, and the skeleton map, the average blood vessel diameter, the average blood vessel density, the proportion of blood vessel length of different diameters, the density of blood vessel layers, and the proportion of vertical blood vessels are calculated to obtain the first blood vessel structure features.

[0010] The second vascular structure characteristics and second epidermal layer thickness were obtained after treatment. Combined with the first epidermal layer thickness and first vascular structure characteristics, the efficacy was analyzed using a logistic regression model to obtain the efficacy evaluation results, which included the V-PDT efficacy prediction probability.

[0011] In one embodiment, the vertical and horizontal gradient components of the tomographic scan are fused, the epidermal-air interface is located using the Dijkstra algorithm, and the thickness of the first epidermal layer is calculated by combining the epidermal-dermal interface, including:

[0012] The vertical and horizontal gradient components of the tomographic scan were calculated using the Sobel operator convolution.

[0013] The weights of the vertical gradient components and the horizontal gradient components are fused to obtain the comprehensive gradient magnitude, and the comprehensive gradient magnitude is then converted into a path weight matrix.

[0014] Initial lateral nodes are obtained from the tomographic scan image. Dijkstra's shortest path algorithm is used to traverse based on the initial lateral nodes and the path weight matrix to generate the shortest path and obtain the skin-air interface.

[0015] The tomographic scan image was flattened by A-line integral projection signal and segmented based on signal intensity threshold to determine the dermal-epidermal interface;

[0016] The thickness of the first epidermal layer is calculated based on the vertical coordinate difference between the epidermal-air interface and the epidermal-dermal interface.

[0017] In one embodiment, the expression for any element of the path weight matrix is:

[0018] w(m,n)=λ1((1-grady(m))+(1-grady(n)))++λ2((1-gradx(m))+(1-gradx(n)))+σ

[0019] Where w(m,n) is any element of the path weight matrix, representing the energy map weight of pixel (m,n) in the tomographic scan image, mgldy(m) and grady(n) are the vertical gradients of pixels m and n respectively, gradx(m) and gradx(n) are the horizontal gradients of pixels m and n respectively, λ1 is the weight coefficient of the vertical gradient, λ2 is the weight coefficient of the horizontal gradient, and σ is a preset constant.

[0020] In one embodiment, adaptive threshold segmentation is performed on the projection image, and median filtering and the Canny operator are combined to extract a binary image of the microvascular network, a vascular contour image, and a skeletal image, including:

[0021] For each pixel in the projection image, calculate the average intensity of the eight neighboring pixels of each pixel, and input the average value into the preset threshold formula to obtain the pixel threshold.

[0022] The pixel intensity of each pixel is compared with the pixel threshold. When the pixel intensity is greater than the pixel threshold, the pixel corresponding to the pixel intensity is marked as a blood vessel region, and the initial binary image is obtained.

[0023] Median filtering is applied to the initial binary image to obtain a binary image of the microvascular network.

[0024] The Canny operator was used to perform boundary detection on the binary map of the microvascular network and extract the outer contour map of the blood vessels.

[0025] Based on the iterative thinning algorithm, the blood vessel edges in the binary image of the microvascular network are transformed into a skeleton image with a single pixel width, thus obtaining the skeleton image.

[0026] In one embodiment, based on the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeletal image, the average blood vessel diameter, average blood vessel density, the proportion of blood vessel lengths of different diameters, the density of blood vessel layers, and the proportion of vertical blood vessels are calculated to obtain the first vascular structure features, including:

[0027] Calculate the ratio of the first pixel of the corresponding blood vessel signal in the binary image of the microvascular network to the second pixel of the corresponding blood vessel signal in the skeleton image, and calculate the average blood vessel diameter.

[0028] Based on the binary map of the microvascular network, the ratio of the area of ​​the vascular region to the total area is calculated to obtain the average vascular density.

[0029] The diameter of each blood vessel is determined based on the skeletal map and the outer contour map of the blood vessels. Then, according to the vessel diameter, each blood vessel is classified according to a preset diameter classification rule. Based on the binary map of the microvascular network and the skeletal map, the proportion of blood vessel length for different diameter categories is calculated using the following formula, thus obtaining the proportion of blood vessel length for different diameters:

[0030]

[0031] Where n is the dimension of the image pixel matrix, (i,j) is the corresponding coordinate position of the pixel in the N×N pixel matrix of the image, and A (i,j) S represents the blood vessel pixels in a binary image of a microvascular network. (i,j) For the skeletal map of blood vessels pixels, X (i,j) c represents the total pixels of the binary image of the vascular network. L For scale;

[0032] Based on the binary image of the microvascular network, the epidermis-dermis interface is divided into superficial and deep layers according to the preset layering rules. The superficial and deep blood vessel pixels are calculated respectively to obtain the blood vessel layering density.

[0033] Based on the outer contour map of blood vessels, the minimum bounding rectangle is used, and vertically running blood vessels are identified by combining preset constraints. The ratio of the corresponding vertically running blood vessel pixel value to the total pixel value in the binary image of the microvascular network is calculated to obtain the proportion of vertical blood vessels.

[0034] The first vascular structural features are obtained by combining the average vessel diameter, average vessel density, the proportion of vessel length of different diameters, the density of vessel layers, and the proportion of vertical vessels.

[0035] In one embodiment, the second vascular structure characteristics and the second epidermal layer thickness after treatment are obtained, and combined with the first epidermal layer thickness and the first vascular structure characteristics, a Logistic regression model is used to analyze the efficacy and obtain the efficacy evaluation results. The efficacy evaluation results include the V-PDT efficacy prediction probability, including:

[0036] One-way ANOVA was used to analyze the first vascular structure characteristics and first epidermal layer thickness before treatment and the second vascular structure characteristics and second epidermal layer thickness after treatment. Parameters associated with V-PDT efficacy were screened out, and parameters with significant efficacy were obtained. The parameters with significant efficacy included epidermal layer thickness, vertical vessel ratio, and vascular layer density.

[0037] The second average vascular density is extracted from the second vascular structure features after treatment, and the difference between the average vascular density and the second average vascular density is calculated to obtain the range of vascular density change.

[0038] By combining parameters that indicate significant therapeutic effects, the magnitude of changes in vascular density, the thickness of the first epidermal layer, the proportion of vertical vessels, and the density of vascular stratification are input into the Logistic regression model, which outputs the predictive probability of V-PDT efficacy. The Logistic regression model is constructed using the forward stepwise regression method.

[0039] In one embodiment, after outputting the predicted probability of V-PDT efficacy, the method further includes:

[0040] Extract the regression coefficients of each input variable from the Logistic regression model, and calculate the corresponding odds ratios using the regression coefficients;

[0041] The statistical significance of the regression coefficients of each input variable was analyzed by Wald test to obtain the probability value of each input variable.

[0042] When the odds ratio and probability value meet the preset conditions, the corresponding input variables are taken as key influencing factors.

[0043] Secondly, this application also provides a system for analyzing the vascular structure of port-wine stains based on OCTA images, including:

[0044] The three-dimensional scanning imaging module is used to perform three-dimensional scanning of the port-wine stain area before treatment using an OCTA system based on the OMAG algorithm, and generate tomographic and projection images.

[0045] The epidermal layer thickness calculation module is used to fuse vertical and horizontal gradient components in a tomographic scan image, locate the epidermal-air interface and epidermal-dermal interface using the Dijkstra algorithm, and calculate the thickness of the first epidermal layer.

[0046] The vascular structure analysis module is used to perform adaptive threshold segmentation on the projection map, and extract the binary map of the microvascular network, the outer contour map of the blood vessel, and the skeleton map by combining median filtering and Canny operator. Based on the binary map of the microvascular network, the outer contour map of the blood vessel, and the skeleton map, the module calculates the average blood vessel diameter, the average blood vessel density, the proportion of blood vessel length of different diameters, the density of blood vessel layering, and the proportion of vertical blood vessels to obtain the first vascular structure features.

[0047] The V-PDT efficacy assessment module is used to obtain the second vascular structure characteristics and the second epidermal layer thickness after treatment. Combined with the first epidermal layer thickness and the first vascular structure characteristics, the efficacy is analyzed through a Logistic regression model to obtain the efficacy assessment results, which include the V-PDT efficacy prediction probability.

[0048] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the first aspect.

[0049] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when processed, implements the steps in the first aspect.

[0050] The aforementioned method and system for analyzing the vascular structure of port-wine stains based on OCTA images utilizes an OCTA system based on the OMAG algorithm to perform three-dimensional scanning of the port-wine stain area before treatment, generating tomographic and projection images. This provides high-resolution three-dimensional images of the skin's microvascular structure, laying a data foundation for subsequent quantitative analysis. Secondly, the tomographic images are fused with vertical and horizontal gradient components. The Dijkstra algorithm is used to locate the epidermal-air interface and the epidermal-dermal interface, calculating the thickness of the first epidermal layer, effectively improving the accuracy of localization and making the measurement of the first epidermal layer thickness more precise. Furthermore, adaptive threshold segmentation is applied to the projection images, combined with median filtering and the Canny operator, to extract the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeletal image. This not only effectively eliminates noise interference but also highlights key features of the vascular structure. Based on these image features, vascular structural features such as average vessel diameter, average vessel density, the proportion of vessel lengths of different diameters, vascular layer density, and the proportion of vertical vessels can be calculated. This allows for quantitative analysis of vascular features from different dimensions, providing strong data support for the pathological study of port-wine stains.

[0051] Finally, by acquiring the post-treatment characteristics of the second vascular structure and the thickness of the second epidermal layer, and combining this with the thickness of the first epidermal layer and the characteristics of the first vascular structure, a logistic regression model was used for efficacy analysis to obtain efficacy evaluation results. This approach not only considers changes in vascular structure before and after treatment but also incorporates epidermal layer thickness, making the efficacy evaluation more comprehensive and accurate. Furthermore, by applying the logistic regression model, multi-factor comprehensive analysis can be used to quantitatively predict the efficacy of V-PDT, providing clinicians with an objective and reliable efficacy evaluation tool. This helps to adjust treatment strategies in advance and improve the success rate of treatment. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart of a method for analyzing the vascular structure of port-wine stains based on OCTA images, provided as an exemplary embodiment of the present invention;

[0054] Figure 2 A schematic diagram of a port-wine stain vascular structure analysis system based on OCTA images, provided as an exemplary embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] In one embodiment, such as Figure 1 As shown, a method for analyzing the vascular structure of port-wine stains based on OCTA images is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0057] S101: The OCTA system based on the OMAG algorithm is used to perform a three-dimensional scan of the port-wine stain area before treatment, generating a tomographic scan image and a projection image.

[0058] Specifically, the OMAG (Optical Microangiography) algorithm is an optical microvascular imaging algorithm that can effectively extract blood flow information and generate high-resolution tomographic and projection maps. A 1310nm superluminescent diode can be used as the light source, and a 3D scan of the port-wine stain lesion area before treatment is performed using an OCTA system based on the OMAG algorithm. The scanning range can be set to 2×2m or 4×4mm, and the scan is repeated 3-5 times to reduce the influence of random noise, obtaining tomographic and projection maps containing blood flow signals. The tomographic map shows the microvascular structure at different depths of the skin, while the projection map provides the vascular distribution throughout the scanned area. Through the tomographic and projection maps, detailed vascular information of the port-wine stain area can be obtained, providing a foundation for subsequent quantitative analysis.

[0059] S102: For the tomographic scan image, fuse the vertical gradient components and the horizontal gradient components, locate the epidermal-air interface and the epidermal-dermal interface using the Dijkstra algorithm, and calculate the thickness of the first epidermal layer.

[0060] Specifically, the thickness of the first epidermal layer is one of the important factors affecting the treatment effect of V-PDT. First, the tomographic scan image can be preprocessed, for example, by performing polynomial fitting and flattening. A quadratic polynomial can be used to eliminate the interference of natural skin surface undulations on interface localization, making subsequent gradient analysis more accurate. Second, the vertical and horizontal gradient components can be extracted from the image. The vertical gradient component reflects the rate of change of the image in the vertical direction, while the horizontal gradient component reflects the rate of change of the image in the horizontal direction. Fusing these two components allows for a more comprehensive capture of the changing characteristics of the epidermal layer boundary. Then, Dijkstra's algorithm can be used, assigning different weights to guide the search direction of the path, enabling it to search for the shortest path along the epidermal layer boundary in the fused gradient image, thereby accurately locating the epidermal-air interface and the epidermal-dermal interface. Finally, the thickness of the first epidermal layer can be calculated based on the distance between the two located interfaces, providing more reliable structural parameters for subsequent efficacy evaluation and helping to more accurately predict treatment effects.

[0061] S103: Adaptive threshold segmentation is performed on the projection map. Median filtering and Canny operator are combined to extract the binary map of the microvascular network, the outer contour map of the blood vessels, and the skeleton map. Based on the binary map of the microvascular network, the outer contour map of the blood vessels, and the skeleton map, the average blood vessel diameter, the average blood vessel density, the proportion of blood vessel length of different diameters, the density of blood vessel layers, and the proportion of vertical blood vessels are calculated to obtain the first blood vessel structure features.

[0062] Specifically, the vascular signal intensity in the projection image is affected by various factors, resulting in uneven distribution. Therefore, an adaptive threshold segmentation method can be used to separate the vascular signal from the background signal in the projection image. This adaptive threshold segmentation dynamically determines the threshold based on the statistical characteristics of the local image, thereby better adapting to the changes in vascular signal intensity in different regions and generating a binary image of the microvascular network. After obtaining the binary image of the microvascular network, median filtering can be used to remove noise points in the image and smooth the vascular edges. Then, the Canny operator can be used to perform edge detection on the processed image to extract the outer contour map of the blood vessels. Furthermore, the Zhang-Suen algorithm can be used to simplify the blood vessels in the binary image of the microvascular network to single-pixel centerlines, obtaining a skeleton map. Illustratively, the Zhang-Suen algorithm is an image thinning algorithm that can simplify the connected regions in the binary image to a single-pixel-width skeleton while preserving the topological structure and key morphological features of the original image. Finally, using the binary image of the microvascular network, the outer contour map of the blood vessels, and the skeleton map, parameters such as the average blood vessel diameter, average blood vessel density, the proportion of blood vessel lengths of different diameters, the density of blood vessel layers, and the proportion of vertical blood vessels can be calculated to obtain the first vascular structure features. This feature can reflect the vascular structure characteristics of the lesion area from multiple perspectives, providing a data basis for subsequent efficacy evaluation.

[0063] S104: Obtain the second vascular structure characteristics and second epidermal layer thickness after treatment, and combine them with the first epidermal layer thickness and first vascular structure characteristics to perform efficacy analysis through a Logistic regression model to obtain efficacy evaluation results. The efficacy evaluation results include the V-PDT efficacy prediction probability.

[0064] Specifically, by comparing the vascular structural features before and after treatment, and combining parameters such as the thickness of the first epidermal layer, the efficacy of vascular-targeted photodynamic therapy (V-PDT) can be evaluated more accurately. Illustratively, after treatment, steps S101 to S103 can be repeated to perform three-dimensional scanning and image processing on the treated port-wine stain area, extracting the post-treatment vascular structural features, including the thickness of the first epidermal layer, average vessel diameter, average vessel density, the proportion of vessel lengths of different diameters, vascular layer density, and the proportion of vertical vessels, thus obtaining the second vascular structural features. Subsequently, based on the first and second vascular structural features, the thickness of the first and second epidermal layers before and after treatment, statistical analysis using a logistic regression model can be performed to calculate the predicted probability of V-PDT efficacy. Furthermore, based on a set threshold, this predicted probability can be divided into different efficacy levels; for example, a probability greater than a certain value indicates effective treatment, while a probability less than a certain value indicates ineffective treatment.

[0065] In the above method, an OCTA system based on the OMAG algorithm is used to perform a three-dimensional scan of the port-wine stain area before treatment, generating tomographic and projection images. This achieves three-dimensional imaging of the vascular structure in the lesion area, comprehensively capturing the morphology and distribution characteristics of blood vessels, and providing reliable data support for subsequent quantitative analysis of vascular structures. Secondly, the vertical and horizontal gradient components of the tomographic images are fused, and the epidermal-air interface and epidermal-dermal interface are located using the Dijkstra algorithm. The thickness of the first epidermal layer is calculated. Gradient fusion enhances the accuracy of interface edge recognition, and combined with the shortest path search characteristic of the Dijkstra algorithm, the measurement of the first epidermal layer thickness is more accurate. Furthermore, adaptive threshold segmentation was applied to the projection map, and median filtering and the Canny operator were used to extract the binary map of the microvascular network, the outer contour map of the blood vessels, and the skeleton map. Adaptive threshold processing solved the problem of insufficient adaptability of traditional fixed thresholds to complex blood vessel distributions, effectively suppressed noise interference and preserved blood vessel edge details, and thus accurately calculated parameters such as average blood vessel diameter and density, realizing multi-dimensional quantitative analysis of the vascular structure of port-wine stains. Finally, by obtaining the second blood vessel structure characteristics and the second epidermal layer thickness after treatment, and combining the first epidermal layer thickness and the first blood vessel structure characteristics, a Logistic regression model was used to analyze the efficacy. An efficacy prediction system including factors such as the proportion of vertical blood vessels and the density of blood vessel layers was constructed, realizing the quantitative prediction of V-PDT efficacy and providing scientific data support for the optimization of clinical treatment plans.

[0066] In one embodiment, for a tomographic scan image, the vertical gradient component and the horizontal gradient component are fused, the epidermal-air interface is located using the Dijkstra algorithm, and the thickness of the first epidermal layer is calculated by combining the epidermal-dermal interface, including:

[0067] The vertical and horizontal gradient components of the tomographic scan were calculated using the Sobel operator convolution.

[0068] The weights of the vertical gradient components and the horizontal gradient components are fused to obtain the comprehensive gradient magnitude, and the comprehensive gradient magnitude is then converted into a path weight matrix.

[0069] Initial lateral nodes are obtained from the tomographic scan image. Dijkstra's shortest path algorithm is used to traverse based on the initial lateral nodes and the path weight matrix to generate the shortest path and obtain the skin-air interface.

[0070] The tomographic scan image was flattened by A-line integral projection signal and segmented based on signal intensity threshold to determine the dermal-epidermal interface;

[0071] The thickness of the first epidermal layer is calculated based on the vertical coordinate difference between the epidermal-air interface and the epidermal-dermal interface.

[0072] Specifically, the Sobel operator is a gradient calculation method that effectively detects edge information in images. Through Sobel convolution, the gradient components in the vertical and horizontal directions of the tomographic image can be calculated separately. These gradient components reflect changes in pixel intensity in the image, helping to detect edge information at the skin-air interface and the skin-dermis interface. Fusing the vertical and horizontal gradient components yields a comprehensive gradient magnitude, which is then transformed into a path weight matrix. The comprehensive gradient magnitude for each pixel is normalized to a value between 0 and 1. Each element in the path weight matrix represents the path weight from one pixel to another. Illustratively, an initial horizontal node in the tomographic image can be chosen as the starting point for path searching. This initial horizontal node is typically located at the left edge of the image, defining the starting position for the path search. Based on the initial horizontal node and the path weight matrix, Dijkstra's shortest path algorithm can be used for path searching. Dijkstra's algorithm is a shortest path search algorithm that finds the path with the minimum path weight from the starting point to the destination. In this embodiment, Dijkstra's algorithm can be used to start from the initial horizontal node, traverse along the path weight matrix, and gradually find the path with the smallest path weight. Finally, the shortest path from the initial horizontal node to the right boundary of the image is generated, which is the skin-air interface.

[0073] In one embodiment, the expression for any element of the path weight matrix is:

[0074] w(m,n)=λ1((1-grady(m))+(1-grady(n)))++λ2((1-gradx(m))+(1-gradx(n)))+σ

[0075] Where w(m,n) is any element of the path weight matrix, representing the energy map weight of pixel (m,n) in the tomographic scan image, mgldy(m) and grady(n) are the vertical gradients of pixels m and n respectively, gradx(m) and gradx(n) are the horizontal gradients of pixels m and n respectively, λ1 is the weight coefficient of the vertical gradient, λ2 is the weight coefficient of the horizontal gradient, and σ is a preset constant.

[0076] After determining the epidermal-air interface, the tomographic scan image can be further processed by A-line integral projection signal flattening to eliminate background noise and inhomogeneities, making the signals of the epidermis and dermis clearer. Specifically, each column of the tomographic scan image can be integrally projected to obtain the A-line signal. Then, the A-line signal is flattened to make the signal intensity variation between the epidermis and dermis more obvious. Based on the flattened A-line signal, a signal intensity threshold can be set, and the location of the epidermal-dermis interface can be determined by threshold segmentation. The signal intensity threshold can be adjusted according to the signal intensity distribution of the actual image to ensure accurate segmentation of the epidermis and dermis. Finally, the thickness of the first epidermal layer can be calculated based on the vertical coordinate difference between the epidermal-air interface and the epidermal-dermis interface.

[0077] In one embodiment, adaptive threshold segmentation is performed on the projection image, and median filtering and the Canny operator are combined to extract a binary image of the microvascular network, a vessel contour image, and a skeleton image, including:

[0078] For each pixel in the projection image, calculate the average intensity of the eight neighboring pixels of each pixel, and input the average value into the preset threshold formula to obtain the pixel threshold.

[0079] The pixel intensity of each pixel is compared with the pixel threshold. When the pixel intensity is greater than the pixel threshold, the pixel corresponding to the pixel intensity is marked as a blood vessel region, and the initial binary image is obtained.

[0080] Median filtering is applied to the initial binary image to obtain a binary image of the microvascular network.

[0081] The Canny operator was used to perform boundary detection on the binary map of the microvascular network and extract the outer contour map of the blood vessels.

[0082] Based on the iterative thinning algorithm, the blood vessel edges in the binary image of the microvascular network are transformed into a skeleton image with a single pixel width, thus obtaining the skeleton image.

[0083] Specifically, the eight neighboring pixels can include pixels above, below, to the left, right, and four diagonally opposite the current pixel. By calculating the average intensity of the neighboring pixels, the local background intensity information of the pixel can be obtained. Then, this average value can be input into a preset threshold formula to obtain the pixel threshold for that pixel. This preset threshold formula can be adjusted according to the statistical characteristics of the actual image, so that each pixel has a dynamic threshold related to its local background intensity, thereby achieving adaptive thresholding. After calculating the pixel threshold for each pixel, its pixel intensity can be compared with its corresponding pixel threshold. If the pixel intensity is greater than the pixel threshold, the pixel is marked as a blood vessel region and assigned a value of 1, i.e., white; otherwise, the pixel is marked as a non-blood vessel region and assigned a value of 0, i.e., black. Through the above process, an initial binary image can be obtained, where blood vessel regions are represented by white and non-blood vessel regions are represented by black.

[0084] The initial binary image may contain some noise points, affecting subsequent vascular structure analysis. Therefore, median filtering can be applied to the initial binary image. Median filtering is a non-linear filtering method that can effectively remove salt-and-pepper noise from the image while preserving edge information. Specifically, for each pixel, the median value of its 3×3 neighborhood pixels can be calculated and assigned to the center pixel. Finally, a median-filtered binary image of the microvascular network is obtained. This image is smoother, and the vascular structure is clearer. Further, the Canny operator can be used for boundary detection to extract the outer contour of the blood vessels. The Canny operator is an edge detection algorithm that can detect strong and weak edges in the image and determine the final edges through non-maximum suppression and threshold detection. Illustratively, the Canny operator first calculates the gradient magnitude and direction of the image, then removes non-edge points through non-maximum suppression, and finally determines strong and weak edges through threshold detection. Through the above process, the outer contour image of the blood vessels can be obtained. This image clearly shows the boundary information of the blood vessels.

[0085] Furthermore, the iterative thinning algorithm is a method for progressively removing edge pixels from an image, transforming the edges of blood vessels into a skeleton of single-pixel width. Illustratively, for each pixel, if two or three of its eight neighboring pixels are background pixels, this pixel is marked as an edge pixel, and the marked edge pixels are set to background, ensuring that removal does not cause the blood vessel to break. This marking and removal operation is repeated until no edge pixels can be removed, resulting in the skeleton image of the blood vessel. This image clearly shows the centerline of the blood vessel, providing an important foundation for subsequent blood vessel feature analysis.

[0086] In one embodiment, based on the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeletal image, the average blood vessel diameter, average blood vessel density, the proportion of blood vessel lengths of different diameters, the density of blood vessel layers, and the proportion of vertical blood vessels are calculated to obtain the first vascular structural features, including:

[0087] Calculate the ratio of the first pixel of the corresponding blood vessel signal in the binary image of the microvascular network to the second pixel of the corresponding blood vessel signal in the skeleton image, and calculate the average blood vessel diameter.

[0088] Based on the binary map of the microvascular network, the ratio of the area of ​​the vascular region to the total area is calculated to obtain the average vascular density.

[0089] The diameter of each blood vessel is determined based on the skeletal map and the outer contour map of the blood vessels. Then, according to the vessel diameter, each blood vessel is classified according to a preset diameter classification rule. Based on the binary map of the microvascular network and the skeletal map, the proportion of blood vessel length for different diameter categories is calculated using the following formula, thus obtaining the proportion of blood vessel length for different diameters:

[0090]

[0091] Where n is the dimension of the image pixel matrix, (i,j) is the corresponding coordinate position of the pixel in the N×N pixel matrix of the image, and A (i,j) S represents the blood vessel pixels in a binary image of a microvascular network. (i,j) For the skeletal map of blood vessels pixels, X (i,j) c represents the total pixels of the binary image of the vascular network. L For scale;

[0092] Based on the binary image of the microvascular network, the epidermis-dermis interface is divided into superficial and deep layers according to the preset layering rules. The superficial and deep blood vessel pixels are calculated respectively to obtain the blood vessel layering density.

[0093] Based on the outer contour map of blood vessels, the minimum bounding rectangle is used, and vertically running blood vessels are identified by combining preset constraints. The ratio of the corresponding vertically running blood vessel pixel value to the total pixel value in the binary image of the microvascular network is calculated to obtain the proportion of vertical blood vessels.

[0094] The first vascular structural features are obtained by combining the average vessel diameter, average vessel density, the proportion of vessel length of different diameters, the density of vessel layers, and the proportion of vertical vessels.

[0095] Specifically, in the binary image of the microvascular network, the vascular region is represented by white pixels, while the vascular centerline in the skeletal image is represented by a single pixel width. The average diameter of the blood vessel can be obtained by calculating the ratio of the number of pixels corresponding to the vascular signal in the two images. The area of ​​the vascular region can be calculated by counting the number of white pixels in the binary image of the microvascular network, while the total area is the total number of pixels in the entire image. By calculating the ratio of the vascular region area to the total area, the average vascular density can be obtained, thus reflecting the distribution of blood vessels in the skin tissue. Furthermore, the diameter of each blood vessel can be determined based on the skeletal image and the outer contour image of the blood vessels. Based on the blood vessel diameter, each blood vessel is classified according to a preset diameter classification rule, for example, dividing every 30μm into intervals such as 10-30μm, 30-60μm, and 60-90μm. Subsequently, each diameter category can be calculated using the above formula. The numerator first obtains the diameter of a single blood vessel by the ratio of the area of ​​the binary image to the length of the skeleton image. Then, the total length of blood vessels in each category interval is calculated, divided by the total scanning area of ​​the binary image, and multiplied by the scale bar to obtain the proportion of blood vessel length of different diameters per unit area.

[0096] Furthermore, considering that V-PDT laser treatment depth is generally limited to the superficial dermis and that V-PDT treatment involves segmental destruction requiring multiple treatment sessions, superficial vessels must already be damaged before deep vessels can absorb sufficient light energy for destruction. Therefore, based on the binary image of the microvascular network, layers can be created at the epidermal-dermal interface according to a preset layering rule. For example, using the epidermal-dermal interface as a baseline, layers can be created in 200μm increments, dividing 0-400μm into the superficial layer and 400-600μm into the deep layer. Subsequently, the corresponding superficial and deep vessel pixels can be calculated from the binary image of the microvascular network, and then divided by the total number of vessel pixels to obtain the superficial / deep density ratio, i.e., the vascular layer density.

[0097] Furthermore, since blood vessels are cylindrical, the projection of vertically running blood vessels onto the skin surface is approximately circular. Illustratively, discontinuous blood vessels can be outlined with a rectangle in the projection image. If this rectangle meets preset constraints, the discontinuous blood vessel can be determined to be approximately circular. These preset constraints include: the length and width of the rectangle should be approximately the same, with a difference not exceeding a preset value; the area ratio of the blood vessel within the rectangle should be greater than a preset area ratio value; and some noise is present in the original image, so some very small white pixels need to be filtered out, meaning the blood vessel ratio must be greater than the preset blood vessel ratio value. After identifying vertically running blood vessels, the ratio of the corresponding vertically running blood vessel pixel value to the total pixel value in the binary image of the microvascular network can be calculated to obtain the vertical blood vessel ratio. Combining the above-mentioned average blood vessel diameter, average blood vessel density, blood vessel length ratio of different diameters, blood vessel layer density, and vertical blood vessel ratio, the first vascular structure feature can be obtained.

[0098] In one embodiment, the second vascular structure characteristics and second epidermal layer thickness after treatment are obtained, and combined with the first epidermal layer thickness and first vascular structure characteristics, a logistic regression model is used for efficacy analysis to obtain efficacy evaluation results. The efficacy evaluation results include the V-PDT efficacy prediction probability, including:

[0099] One-way ANOVA was used to analyze the first vascular structure characteristics and first epidermal layer thickness before treatment and the second vascular structure characteristics and second epidermal layer thickness after treatment. Parameters associated with V-PDT efficacy were screened out, and parameters with significant efficacy were obtained. The parameters with significant efficacy included epidermal layer thickness, vertical vessel ratio, and vascular layer density.

[0100] The second average vascular density is extracted from the second vascular structure features after treatment, and the difference between the average vascular density and the second average vascular density is calculated to obtain the range of vascular density change.

[0101] By combining parameters that indicate significant therapeutic effects, the magnitude of changes in vascular density, the thickness of the first epidermal layer, the proportion of vertical vessels, and the density of vascular stratification are input into the Logistic regression model, which outputs the predictive probability of V-PDT efficacy. The Logistic regression model is constructed using the forward stepwise regression method.

[0102] Specifically, a one-way ANOVA can be performed on the first vascular structure characteristics and first epidermal layer thickness before treatment, and the second vascular structure characteristics and second epidermal layer thickness after treatment. One-way ANOVA is a statistical method used to compare the differences in means among multiple groups, and can test whether there are significant differences in the means of a certain parameter between different groups. In this embodiment, the correlation between each parameter and the efficacy can be determined by calculating the statistic F and the corresponding probability P value. If the P value is less than a pre-set significance level, then the parameter can be considered to have a significant association with the efficacy of V-PDT. For example, for epidermal layer thickness, the mean epidermal layer thickness of different efficacy groups, such as the effective group and the ineffective group, can be calculated, and the variance analysis can be used to determine whether there is a significant difference between the means of the two groups. After analysis, significant efficacy parameters such as epidermal layer thickness, vertical vessel ratio, and vascular layer density can be screened out; that is, these parameters have significant differences between different efficacy groups and have an important impact on the efficacy of V-PDT.

[0103] The difference between the pre-treatment average vascular density and the post-treatment average vascular density can then be calculated to obtain the magnitude of vascular density change. This magnitude of change reflects the alteration of vascular density before and after treatment and is a crucial indicator for evaluating the efficacy of V-PDT. Based on the selected significant efficacy parameters and the calculated magnitude of vascular density change, a logistic regression model can be constructed using forward stepwise regression. Forward stepwise regression is a variable selection method that starts with a model without independent variables and gradually introduces independent variables, introducing the one that contributes most to the model each time, until no independent variable significantly improves the model. By using forward stepwise regression, it ensures that the variables included in the model significantly contribute to efficacy prediction, thereby improving the model's accuracy and reliability. The logistic regression model is a generalized linear model that can predict the probability of V-PDT efficacy, i.e., the probability of improvement in the vascular structure of the lesion area after V-PDT treatment, i.e., the V-PDT efficacy prediction probability. This probability not only provides an objective basis for efficacy evaluation but also helps clinicians adjust treatment plans in advance, improving the success rate of treatment.

[0104] In one embodiment, after outputting the predicted probability of V-PDT efficacy, the method further includes:

[0105] Extract the regression coefficients of each input variable from the Logistic regression model, and calculate the corresponding odds ratios using the regression coefficients;

[0106] The statistical significance of the regression coefficients of each input variable was analyzed by Wald test to obtain the probability value of each input variable.

[0107] When the odds ratio and probability value meet the preset conditions, the corresponding input variables are taken as key influencing factors.

[0108] Specifically, regression coefficients for each input variable can be extracted from the logistic regression model. These coefficients reflect the degree of influence of each input variable on the predicted probability of therapeutic efficacy. Then, the odds ratio (OR) is calculated using the regression coefficients. The OR is an important indicator of the strength of the association between the input variable and the therapeutic efficacy. The Wald test is a statistical test used to determine whether the regression coefficient is significantly different from 0. The Wald test yields the probability value for each input variable. The smaller the probability value, the more significant the impact of the input variable on the therapeutic efficacy. For example, when the probability value is less than 0.05, the input variable can be considered to have a significant impact on the therapeutic efficacy. Subsequently, the OR and probability values ​​can be used to further determine whether the input variable is a key influencing factor. A preset condition is an OR greater than 1 and a P-value less than 0.05. Input variables that meet this preset condition can be considered key factors affecting the efficacy of V-PDT. For example, if variables such as epidermal thickness, vessel diameter, and vessel density are identified as key influencing factors, in clinical practice, doctors can use these key factors to formulate more precise treatment plans and predict the therapeutic efficacy for patients. Furthermore, these key factors can be used to further optimize V-PDT treatment parameters and improve treatment outcomes.

[0109] like Figure 2 As shown, based on the same inventive concept, this application also provides a port-wine stain vascular structure analysis system 200 based on OCTA images for implementing the aforementioned method for analyzing the vascular structure of port-wine stains based on OCTA images. The solution provided by this system is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the port-wine stain vascular structure analysis system based on OCTA images provided below can be found in the above-described limitations of the method for analyzing the vascular structure of port-wine stains based on OCTA images, and will not be repeated here. The system includes:

[0110] The three-dimensional scanning imaging module is used to perform three-dimensional scanning of the port-wine stain area before treatment using an OCTA system based on the OMAG algorithm, and generate tomographic and projection images.

[0111] The epidermal layer thickness calculation module is used to fuse vertical and horizontal gradient components in a tomographic scan image, locate the epidermal-air interface and epidermal-dermal interface using the Dijkstra algorithm, and calculate the thickness of the first epidermal layer.

[0112] The vascular structure analysis module is used to perform adaptive threshold segmentation on the projection map, and extract the binary map of the microvascular network, the outer contour map of the blood vessel, and the skeleton map by combining median filtering and Canny operator. Based on the binary map of the microvascular network, the outer contour map of the blood vessel, and the skeleton map, the module calculates the average blood vessel diameter, the average blood vessel density, the proportion of blood vessel length of different diameters, the density of blood vessel layering, and the proportion of vertical blood vessels to obtain the first vascular structure features.

[0113] The V-PDT efficacy assessment module is used to obtain the second vascular structure characteristics and the second epidermal layer thickness after treatment. Combined with the first epidermal layer thickness and the first vascular structure characteristics, the efficacy is analyzed through a Logistic regression model to obtain the efficacy assessment results, which include the V-PDT efficacy prediction probability.

[0114] In one exemplary embodiment, the present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method for analyzing the vascular structure of port-wine stains based on OCTA images according to this application. A multi-core processor is preferred to improve the parallel processing capability of the system. The memory provides sufficient temporary storage space to support program execution and data processing. The memory capacity should be large enough to accommodate a large amount of information and computational tasks.

[0115] In one exemplary embodiment, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for analyzing the vascular structure of port-wine stains based on OCTA images according to this application.

[0116] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for analyzing the vascular structure of port-wine stains based on OCTA images, characterized in that, The method includes: The OCTA system based on the OMAG algorithm was used to perform a three-dimensional scan of the port-wine stain area before treatment, generating a tomographic scan and a projection image. For the tomographic scan image, the vertical gradient component and the horizontal gradient component are fused, and the epidermal-air interface and the epidermal-dermal interface are located using the Dijkstra algorithm to calculate the thickness of the first epidermal layer. Adaptive threshold segmentation is performed on the projection image, and median filtering and Canny operator are combined to extract the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeleton image. Based on the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeleton image, the average blood vessel diameter, the average blood vessel density, the proportion of blood vessel length of different diameters, the blood vessel layer density, and the proportion of vertical blood vessels are calculated to obtain the first blood vessel structure features. The second vascular structure features and second epidermal layer thickness after treatment are obtained, and combined with the first epidermal layer thickness and the first vascular structure features, the efficacy is analyzed by a Logistic regression model to obtain the efficacy evaluation results, which include the V-PDT efficacy prediction probability.

2. The method according to claim 1, characterized in that, The tomographic scan image is fused with vertical and horizontal gradient components. The epidermal-air interface is located using Dijkstra's algorithm, and the thickness of the first epidermal layer is calculated by combining the epidermal-dermal interface data. This includes: The vertical gradient component and the horizontal gradient component of the tomographic scan are calculated by convolution using the Sobel operator; The weights of the vertical gradient components and the horizontal gradient components are fused to obtain a comprehensive gradient magnitude, and the comprehensive gradient magnitude is then converted into a path weight matrix. In the tomographic scan image, an initial lateral node is obtained. Using Dijkstra's shortest path algorithm, the path is traversed based on the initial lateral node and the path weight matrix to generate the shortest path, thus obtaining the skin-air interface. The dermal-epidermal interface is determined by flattening the A-line integral projection signal of the tomographic scan image and segmenting it based on the signal intensity threshold. The thickness of the first epidermal layer is calculated based on the vertical coordinate difference between the epidermal-air interface and the epidermal-dermal interface.

3. The method according to claim 2, characterized in that, The expression for any element of the path weight matrix is: w(m,n)=λ1((1-grady(m))+(1-grady(n)))++λ2((1-gradx(m))+(1-gradx(n)))+σ Where w(m,n) is any element of the path weight matrix, representing the energy map weight of pixel (m,n) in the tomographic scan image, mgldy(m) and grady(n) are the vertical gradients of pixels m and n respectively, gradx(m) and gradx(n) are the horizontal gradients of pixels m and n respectively, λ1 is the weight coefficient of the vertical gradient, λ2 is the weight coefficient of the horizontal gradient, and σ is a preset constant.

4. The method according to claim 1, characterized in that, The adaptive threshold segmentation of the projection image, combined with median filtering and the Canny operator, extracts the binary image of the microvascular network, the outer contour image of the blood vessels, and the skeleton image, including: For each pixel in the projection image, calculate the average intensity of the eight neighboring pixels of each pixel, and input the average value into a preset threshold formula to obtain the pixel threshold. The pixel intensity of each pixel is compared with the pixel threshold. When the pixel intensity is greater than the pixel threshold, the pixel corresponding to the pixel intensity is marked as a blood vessel region to obtain an initial binary image. Median filtering is applied to the initial binary image to obtain the microvascular network binary image; The Canny operator is used to perform boundary detection on the binary image of the microvascular network to extract the outer contour image of the blood vessels; The iterative thinning algorithm is used to transform the blood vessel edges in the binary image of the microvascular network into a skeleton image with a single pixel width, thereby obtaining the skeleton image.

5. The method according to claim 4, characterized in that, The first vascular structure feature is obtained by calculating the average vessel diameter, average vessel density, proportion of vessel lengths of different diameters, vascular layer density, and proportion of vertical vessels based on the binary image of the microvascular network, the outer contour image of the vessels, and the skeletal image. This includes: The average blood vessel diameter is calculated by comparing the ratio of the first pixel point corresponding to the blood vessel signal in the binary image of the microvascular network to the second pixel point corresponding to the blood vessel signal in the skeletal image. Based on the binary map of the microvascular network, the ratio of the area of ​​the vascular region to the total area is calculated to obtain the average vascular density. The diameter of each blood vessel is determined based on the skeletal diagram and the outer contour diagram of the blood vessels. Based on the blood vessel diameter, each blood vessel is classified according to a preset diameter classification rule. The proportion of blood vessel length for different diameter categories is calculated using the following formula based on the binary diagram of the microvascular network and the skeletal diagram, thus obtaining the proportion of blood vessel length for different diameters: Where n is the dimension of the image pixel matrix, (i,j) is the corresponding coordinate position of the pixel in the N×N pixel matrix of the image, and A (i,j) S represents the blood vessel pixels in the binary image of the microvascular network. (i,j) For the vascular pixels in the skeletal map, X (i,j) c represents the total number of pixels in the binary image of the vascular network. L For scale; Based on the binary image of the microvascular network, the epidermis-dermis interface is divided into a superficial layer and a deep layer according to a preset layering rule. The superficial blood vessel pixels and deep blood vessel pixels corresponding to the superficial layer and deep layer are calculated respectively to obtain the blood vessel layering density. Based on the outer contour map of the blood vessels, the minimum bounding rectangle is used, and the vertically running blood vessels are identified by combining the preset constraints. The ratio of the pixel value of the vertically running blood vessel to the total pixel value in the binary image of the microvascular network is calculated to obtain the proportion of the vertical blood vessels. The first vascular structural feature is obtained by combining the average vessel diameter, the average vessel density, the proportion of vessel lengths of different diameters, the vascular layer density, and the proportion of vertical vessels.

6. The method according to claim 1, characterized in that, The process involves acquiring the second vascular structure characteristics and second epidermal layer thickness after treatment, and combining these with the first epidermal layer thickness and the first vascular structure characteristics. A logistic regression model is then used to analyze the treatment efficacy, yielding efficacy evaluation results, including: One-way ANOVA was used to analyze the first vascular structure characteristics and the first epidermal layer thickness before treatment and the second vascular structure characteristics and the second epidermal layer thickness after treatment. Parameters associated with the efficacy of V-PDT were screened to obtain parameters with significant efficacy, including epidermal layer thickness, the proportion of vertical vessels, and the density of vascular layering. The second average vascular density is extracted from the second vascular structure features after treatment, and the difference between the average vascular density and the second average vascular density is calculated to obtain the vascular density change range. Combining the aforementioned significant efficacy parameters, the magnitude of the change in vascular density, the thickness of the first epidermal layer, the proportion of vertical blood vessels, and the density of vascular layering are input into the Logistic regression model, and the predicted probability of V-PDT efficacy is output, wherein the Logistic regression model is constructed using the forward stepwise regression method.

7. The method according to claim 6, characterized in that, After outputting the predicted probability of V-PDT efficacy, the method further includes: The regression coefficients of each input variable are extracted from the Logistic regression model, and the corresponding odds ratios are calculated using the regression coefficients. The regression coefficients of each input variable were statistically significant using the Wald test to obtain the probability values ​​of each input variable. When the odds ratio and the probability value meet preset conditions, the corresponding input variable is taken as a key influencing factor.

8. A system for analyzing the vascular structure of port-wine stains based on OCTA images, characterized in that, The system includes: The three-dimensional scanning imaging module is used to perform three-dimensional scanning of the port-wine stain area before treatment using an OCTA system based on the OMAG algorithm, and generate tomographic and projection images. The epidermal layer thickness calculation module is used to fuse the vertical gradient component and the horizontal gradient component in the tomographic scan image, locate the epidermal-air interface and the epidermal-dermal interface using the Dijkstra algorithm, and calculate the thickness of the first epidermal layer. The vascular structure analysis module is used to perform adaptive threshold segmentation on the projection map, and extract the microvascular network binary map, vascular outer contour map and skeleton map by combining median filtering and Canny operator. Based on the microvascular network binary map, the vascular outer contour map and the skeleton map, the module calculates the average vascular diameter, average vascular density, the proportion of vascular length of different diameters, vascular layer density and vertical vascular proportion to obtain the first vascular structure feature. The V-PDT efficacy assessment module is used to obtain the second vascular structure characteristics and the second epidermal layer thickness after treatment, and combine the first epidermal layer thickness and the first vascular structure characteristics to perform efficacy analysis through a Logistic regression model to obtain efficacy assessment results. The efficacy assessment results include the V-PDT efficacy prediction probability.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images