Orthopedic postoperative scar recovery dynamic evaluation system based on deep learning

Abstract

The present application relates to the technical field of deep learning, and particularly relates to a kind of dynamic evaluation system of plastic surgery postoperative scar recovery based on deep learning, comprising: the gray image of dermoscope at different sampling moments is obtained, and scar area and microvessel area around scar are obtained by division;According to the area change of microvessel texture distribution feature in microvessel area scar area, determine recovery state index;When hyperplastic scar appears, determine the area and gray scale of hyperplastic scar in scar area, according to area feature and gray scale feature, determine the proliferation degree of hyperplastic scar;According to the proliferation degree in different gray images, adjust the recovery state index, obtain recovery evaluation coefficient, and determine recovery evaluation result according to recovery evaluation coefficient under different sampling moments.The present application can evaluate in time by combining overall recovery state and the proliferation of pathological scar area, effectively improve the accuracy and reliability of the evaluation result of scar recovery.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep learning, and in particular to a plastic surgery postoperative scar recovery dynamic evaluation system based on deep learning. BACKGROUND

[0002] Scars are an inevitable product of the body's wound repair process. Traditional scar evaluation methods mainly rely on the clinical experience of doctors or dermoscopy methods. However, these methods cannot dynamically and comprehensively reflect the recovery of scars. With the continuous progress of science and technology, such as the development of deep learning combined with dermoscopy imaging technology, new technologies can analyze the recovery of scars by capturing wound surface images, thereby comprehensively and dynamically analyzing the recovery of scars.

[0003] Currently, there are two types of scars after plastic surgery: physiological scars and pathological scars. Physiological scars are scars that can heal normally after surgery and do not have excessive proliferation or defects. However, during the wound repair process, factors such as extracellular matrix metabolism imbalance, abnormal activation of fibroblasts, and inflammatory response can lead to the formation of pathological scars. In the existing method for evaluating the recovery of scars after plastic surgery in patients, the image analysis-based method identifies scars and performs scar labeling. However, this method cannot timely evaluate the overall recovery state and the proliferation of pathological scars, resulting in biased evaluation results of the recovery of scars after plastic surgery in patients. SUMMARY

[0004] To solve the technical problem that the overall recovery state and the proliferation of pathological scars cannot be timely evaluated in the related art, resulting in biased evaluation results of the recovery of scars after plastic surgery in patients, the present application provides a plastic surgery postoperative scar recovery dynamic evaluation system based on deep learning, and the technical solution adopted is as follows:

[0005] The present application provides a plastic surgery postoperative scar recovery dynamic evaluation system based on deep learning, which comprises:

[0006] An acquisition module is configured to acquire a gray-scale image of a dermoscope at different sampling times in a scar recovery stage, perform edge detection on the gray-scale image to obtain an edge image, and divide the edge image to obtain a scar area and a microvessel area around the scar.

[0007] A recovery analysis module is configured to determine a blood vessel distribution feature index of the microvessel area according to the microvessel texture density and distribution uniformity in the microvessel area, and determine a recovery state index at the sampling time according to the change of the blood vessel distribution feature index in the adjacent gray-scale images in time sequence and the area change of the scar area.

[0008] The proliferation analysis module is configured to determine whether a hypertrophic scar occurs according to a texture change of the scar area in different gray images, and determine an area and a gray scale of the hypertrophic scar in the scar area when the hypertrophic scar occurs, and determine a proliferation degree of the hypertrophic scar according to the area feature and the gray scale feature.

[0009] The evaluation module is configured to adjust a recovery state index according to the proliferation degree in different gray images to obtain a recovery evaluation coefficient, and determine a recovery evaluation result according to the recovery evaluation coefficient at different sampling moments.

[0010] Further, edge detection is performed on the gray image to obtain an edge image, including:

[0011] The edge detection is performed on the gray image based on a Canny edge detection algorithm to obtain edge pixel points.

[0012] The gray image is binarized according to the edge pixel points to obtain the edge image, wherein the edge pixel points correspond to a pixel value of 1, and other pixel points correspond to a pixel value of 0.

[0013] Further, a scar area and a microvessel area around the scar are divided from the edge image, including:

[0014] The scar area and the microvessel area around the scar are identified based on a pre-trained image feature recognition model.

[0015] Further, a blood vessel distribution feature index of the microvessel area is determined according to a microvessel texture density and a distribution uniformity in the microvessel area, including:

[0016] The number ratio of the edge pixel points in the microvessel area is calculated as the microvessel texture density;

[0017] The distance between any two edge pixel points and the number distribution of the edge pixel points in different sub-areas in the microvessel area are determined to obtain a distribution uniformity index;

[0018] The product value of the microvessel texture density and the distribution uniformity index is calculated and normalized as the blood vessel distribution feature index of the microvessel area.

[0019] Further, the distribution uniformity index is determined according to the distance between any two edge pixel points and the number distribution of the edge pixel points in different sub-areas in the microvessel area, including:

[0020] The microvessel area is evenly divided into sub-areas of a preset size, and the number variance of the edge pixel points in each sub-area is determined as a number distribution dispersion coefficient;

[0021] The product value of the distance mean and the distance variance is normalized as a distance distribution discrete coefficient;

[0022] The sum value of the quantity distribution discrete coefficient and the distance distribution discrete coefficient is taken as a total discrete index, and the reciprocal of the total discrete index is normalized as a distribution uniformity index.

[0023] Further, the recovery state index at the sampling moment is determined according to the change of the blood vessel distribution feature index in the adjacent gray images in time sequence and the area change of the scar area, and includes:

[0024] The difference value of the blood vessel distribution feature index in the gray images between any sampling moment and the previous sampling moment in time sequence is normalized as a blood vessel recovery coefficient at the corresponding sampling moment;

[0025] The area difference value of the scar area between the previous sampling moment and the corresponding sampling moment in time sequence is normalized to obtain a scar recovery coefficient;

[0026] The sum value of the blood vessel recovery coefficient and the scar recovery coefficient is normalized as a recovery state index.

[0027] Further, whether the hypertrophic scar appears is determined according to the texture change of the scar area in different gray images, and includes:

[0028] The skeleton of the scar area in different gray images is extracted to determine the scar skeleton pixel points in each gray image;

[0029] The scar skeleton pixel points in the latter frame and not in the former frame gray image in the adjacent two frames are taken as the proliferation pixel points;

[0030] The ratio of the number of the proliferation pixel points to the number of the scar skeleton pixel points in the corresponding latter frame gray image is taken as a proliferation coefficient;

[0031] When the proliferation coefficient is greater than a preset proliferation threshold, it is determined that the hypertrophic scar appears, otherwise, it is determined that the hypertrophic scar does not appear.

[0032] Further, the proliferation degree of the hypertrophic scar is determined according to the area feature and the gray feature, and includes:

[0033] The proliferation coefficient is taken as the area feature; the gray mean value of all the proliferation pixel points is calculated, and the reciprocal of the gray mean value is normalized as a gray feature index;

[0034] The product value of the proliferation coefficient and the gray feature index is normalized as the proliferation degree.

[0035] Further, according to the proliferation degree in different gray images, the recovery state index is adjusted to obtain a recovery evaluation coefficient, including:

[0036] The product value of the proliferation degree and the preset influence weight is taken as the adjustment index;

[0037] The difference value between the recovery state index and the adjustment index is calculated as the recovery evaluation coefficient.

[0038] Further, according to the recovery evaluation coefficient at different sampling moments, a recovery evaluation result is determined, including:

[0039] The average value of the recovery evaluation coefficients at all sampling moments is calculated to obtain a total recovery index;

[0040] When the value of the total recovery index is greater than 0.7, the recovery evaluation result is determined to be excellent, otherwise, the recovery evaluation result is determined to be general.

[0041] The present application has the following beneficial effects:

[0042] The embodiment of the present application obtains the gray images of the dermatoscope at different sampling moments in the scar recovery stage, determines the scar area and the microvessel area, and then performs classification analysis, so that the processing mode is more accurate and reliable compared with directly analyzing the whole scar, and the analysis result obtained is more accurate; in the recovery analysis module, the blood vessel distribution characteristic index is determined according to the microvessel texture density and the distribution uniformity in the microvessel area, and the recovery state index is determined according to the change of the blood vessel distribution characteristic index in time sequence and the area change of the scar area, so that the recovery state index can be analyzed in two dimensions of the microvessel texture feature and the scar area feature, and the accuracy of the recovery state index is higher; in the proliferation analysis module, the proliferation degree of the hypertrophic scar is determined by analyzing the area and the gray state of the hypertrophic scar; and in the evaluation module, the recovery state index is adjusted according to the proliferation degree, and the recovery evaluation result is comprehensively determined. In summary, the present application can timely evaluate the overall recovery state and the proliferation condition of the pathological scar area, and effectively improves the accuracy and reliability of the evaluation result of the scar recovery. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0044] Figure 1A structure diagram of a deep learning-based postoperative scar recovery dynamic evaluation system for an embodiment of the present application is provided.

[0045] Figure 2 A gray-scale image schematic diagram of a microvessel region for an embodiment of the present application is provided.

[0046] Figure 3 An edge image schematic diagram of a microvessel region for an embodiment of the present application is provided.

[0047] Figure 4 A gray-scale image schematic diagram of a scar region for an embodiment of the present application is provided.

[0048] Figure 5 An edge image schematic diagram of a scar region for an embodiment of the present application is provided.

[0049] Figure 6 An edge image schematic diagram of a microvessel region under normal conditions for an embodiment of the present application is provided. DETAILED DESCRIPTION

[0050] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined purposes, the following describes in detail the specific implementation, structure, features and effects of a deep learning-based postoperative scar recovery dynamic evaluation system according to the present application, in combination with the preferred embodiments and the drawings. In the following description, different “one embodiment” or “another embodiment” do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0052] The following specifically describes the specific scheme of a deep learning-based postoperative scar recovery dynamic evaluation system according to the present application in combination with the drawings.

[0053] Please refer to Figure 1 , which shows a structure diagram of a deep learning-based postoperative scar recovery dynamic evaluation system for an embodiment of the present application. The system includes an acquisition module 101, a recovery analysis module 102, a hyperplasia analysis module 103 and an evaluation module 104. The specific analysis is carried out subsequently.

[0054] The acquisition module 101 is used to acquire the gray-scale image of the dermatoscope at different sampling times in the scar recovery stage, perform edge detection on the gray-scale image to obtain an edge image, and divide the edge image to obtain the scar region and the microvessel region around the scar.

[0055] Scars are an inevitable product of the body's wound repair process. Traditional scar assessment methods mainly rely on the clinical experience of doctors or dermoscopy and other methods. However, these methods cannot dynamically and comprehensively reflect the recovery of the scar. With the continuous progress of science and technology, such as the development of 3D imaging technology, new technologies can capture wound surface images for analysis, thereby comprehensively and dynamically analyzing the recovery of the scar.

[0056] Currently, there are two types of scars after plastic surgery, namely physiological scars and pathological scars. Physiological scars are scars formed by surgery that can heal normally and have no excessive proliferation or defects. During the wound repair process, if factors such as extracellular matrix metabolism imbalance, abnormal activation of fibroblasts, and inflammatory response occur, pathological scars will be produced. In the process of evaluating the recovery of scars after plastic surgery in patients, the existing method identifies scars based on image analysis and performs scar labeling, but cannot timely evaluate the overall recovery state and the proliferation of pathological scars, resulting in deviations in the evaluation results of the recovery of scars after plastic surgery in patients.

[0057] To solve the above problems, in the embodiments of the present application, a dermoscope is used to collect a skin surface image, and then image preprocessing and image grayscale processing are performed to obtain a grayscale image. Recovery analysis and proliferation analysis are performed based on the texture features in the grayscale image, thereby accurately realizing the evaluation of the recovery of scars after plastic surgery in patients.

[0058] In the embodiments of the present application, edge detection is performed on the grayscale image to determine an edge image. In some embodiments of the present application, the edge detection method can specifically use a Canny edge detection algorithm, that is, the grayscale image is edge detected based on the Canny edge detection algorithm to obtain edge pixel points. The grayscale image is binarized based on the edge pixel points to obtain the edge image, wherein the edge pixel points correspond to a pixel value of 1, and other pixel points correspond to a pixel value of 0.

[0059] The Canny edge detection algorithm can realize edge detection of the image to determine the edge pixel points, and the edge image is obtained by binarizing the edge pixel points. Specifically, in some embodiments of the present application, the pixel point corresponding to the pixel value of 1 is the edge pixel point, and the grayscale value in the edge image is 255, while the non-edge pixel point has a pixel value of 0 and a corresponding grayscale value of 0. Thus, binary image analysis is realized, which facilitates subsequent analysis based on edge features.

[0060] Since the scar area mainly includes two parts, one is the scar area to be recovered after the surgical incision is sutured, and the other is the microvessel area affected around the scar area, since the scar presents the condition of blood vessel expansion and distribution disorder in the dermoscopy image. The microvessels of normal skin are uniformly distributed, and there is no abnormal blood vessel expansion and the like, therefore, in the embodiment of the application, the area of blood vessel expansion and distribution disorder is called the microvessel area, which is mainly distributed around the scar area and presents an irregular state, and it is necessary to analyze respectively in order to obtain more accurate evaluation results.

[0061] Referring to Figures 2 to 5 , Figure 2 The gray scale image schematic diagram of the microvessel area provided by an embodiment of the application, Figure 3 The edge image schematic diagram of the microvessel area provided by an embodiment of the application, Figure 4 The gray scale image schematic diagram of the scar area provided by an embodiment of the application, Figure 5 The edge image schematic diagram of the scar area provided by an embodiment of the application. By Figures 2 to 5 It can be known that the scar area presents an area obviously with knife edge texture, and the microvessel area is an abnormal area generated by the influence of factors such as inflammation reaction, and is usually located around the scar area.

[0062] The scar area and the microvessel area around the scar area are divided from the edge image, including: based on the pre-trained image feature recognition model, the scar area and the microvessel area around the scar area are recognized.

[0063] Since the texture features of the scar area and the microvessel area are obvious, thereby, the pre-trained image feature recognition model can be obtained by directly training a big data model based on the labeled image, and then the efficient recognition of the scar area and the microvessel area is realized based on the model, which is well known to the related technical personnel in the art, and will not be described here.

[0064] The recovery analysis module 102 is used for determining the blood vessel distribution feature index of the microvessel area according to the microvessel texture density and the distribution uniformity in the microvessel area, and determining the recovery state index at the sampling moment according to the change of the blood vessel distribution feature index in the time sequence adjacent gray scale images and the area change of the scar area.

[0065] In the embodiment of the application, the change of the microvessel texture distribution feature in the microvessel area and the change of the scar itself can be analyzed by the dermoscope to analyze the recovery state, the blood vessel distribution feature index in the embodiment of the application represents the texture distribution of the microvessel area, and the gray scale state index represents the recovery at the corresponding sampling moment.

[0066] Further, in some embodiments of the present application, according to the microvessel texture density and distribution uniformity in the microvessel region, the blood vessel distribution characteristic index of the microvessel region is determined, including: calculating the number proportion of the edge pixel points in the microvessel region as the microvessel texture density; according to the distance between any two edge pixel points and the number distribution of the edge pixel points in different sub-regions in the microvessel region, the distribution uniformity index is determined; the product value of the microvessel texture density and the distribution uniformity index is calculated and normalized as the blood vessel distribution characteristic index of the microvessel region.

[0067] Referring to Figure 6 , Figure 6 The edge image schematic diagram of the microvessel region under normal circumstances provided by an embodiment of the present application; the edge image of the microvessel region under abnormal circumstances is shown in FIG. 2. Figure 6 Compared with Figure 3 It can be obviously known that the microvessel texture distribution under normal circumstances is more average, and the blood vessel density under abnormal circumstances is also small due to the uneven blood vessel distribution caused by the scar wound.

[0068] In the embodiment of the present application, the number proportion of the edge pixel points in the microvessel region is calculated as the microvessel texture density, and the greater the microvessel texture density, the more corresponding to the blood vessel distribution characteristic under normal circumstances.

[0069] In the embodiment of the present application, for the analysis of the distribution uniformity, the distance and number distribution of the edge pixel points in different sub-regions need to be judged, that is, the microvessel region is divided into different sub-regions, and then the number distribution of the edge pixel points in each sub-region is analyzed for uniformity.

[0070] According to the distance between any two edge pixel points and the number distribution of the edge pixel points in different sub-regions in the microvessel region, the distribution uniformity index is determined, including: the microvessel region is divided into sub-regions of a preset size, the number variance of the edge pixel points in each sub-region is determined as the number distribution dispersion coefficient; the distance mean and distance variance between any two edge pixel points are calculated, the product value of the distance mean and the distance variance is normalized as the distance distribution dispersion coefficient; the sum of the number distribution dispersion coefficient and the distance distribution dispersion coefficient is taken as the total dispersion index, the inverse of the total dispersion index is calculated and normalized as the distribution uniformity index.

[0071] Wherein, the preset size is the size of the sub-region, in the embodiment of the present application, the preset size can be determined as 5x5 size, that is, the microvessel region is divided into different 5x5 size sub-regions, of course, it can also be adjusted to 7x7 size and other different forms according to the actual situation, which is not limited.

[0072] Discrete analysis is conducted on the number of edge pixel points in each sub-region, wherein the number variance of the edge pixel points in all sub-regions is calculated as a number distribution discrete coefficient, and the greater the value of the number distribution discrete coefficient, the more uneven the number distribution of the edge pixel points in different sub-regions, that is, the lower the distribution uniformity.

[0073] The distance of different edge pixel points is analyzed, which is divided into two categories of distance mean and distance variance, the distance mean represents the density feature, the greater the distance mean, the smaller the density distribution, and the distance variance represents the distance uniformity feature, the greater the distance variance, the more uneven the distance distribution, and thus the product value of the distance mean and the distance variance is normalized as a distance distribution discrete coefficient.

[0074] Since the greater the values of the number distribution discrete coefficient and the distance distribution discrete coefficient, the lower the overall distribution uniformity, in the embodiments of the present application, the sum of the number distribution discrete coefficient and the distance distribution discrete coefficient is directly calculated as a total discrete index, the reciprocal of the total discrete index is calculated and normalized as a distribution uniformity index, so that the greater the value of the distribution uniformity index, the more in line with the uniform distribution feature.

[0075] The product value of the microvessel texture density and the distribution uniformity index is calculated and normalized as a blood vessel distribution feature index of the microvessel region, and the greater the value of the blood vessel distribution feature index, the more the overall microvessel distribution tends to be normal.

[0076] Further, in some embodiments of the present application, the recovery state index at the sampling time is determined according to the change of the blood vessel distribution feature index in the time-series adjacent gray-scale images and the area change of the scar region, including: the difference between the blood vessel distribution feature index of any sampling time and the previous sampling time in the gray-scale image is normalized as a blood vessel recovery coefficient at the corresponding sampling time; the area difference of the scar region between the previous sampling time and the corresponding sampling time in the time series is calculated and normalized to obtain a scar recovery coefficient; and the sum of the blood vessel recovery coefficient and the scar recovery coefficient is normalized as the recovery state index.

[0077] The blood vessel recovery coefficient represents the change in the degree of microvessel tending to be normal between two adjacent sampling times, and the greater the value of the blood vessel recovery coefficient, the greater the value of the blood vessel distribution feature index, that is, the more the overall microvessel distribution tends to be normal, and the more obvious the recovery effect.

[0078] In the embodiment of the present application, the area change of the scar area in time sequence needs to be analyzed. With healing, the scar gradually becomes smaller until scabbing, so the area of the scar area is gradually reduced. The area difference between two adjacent sampling time points is taken as the scar recovery coefficient. The greater the scar recovery coefficient value is, the faster the scar recovers at the two adjacent sampling time points, and the more obvious the recovery effect is.

[0079] In summary, the sum of the blood vessel recovery coefficient and the scar recovery coefficient is calculated and normalized as the recovery state index.

[0080] The hyperplasia analysis module 103 is configured to determine whether hyperplastic scar appears according to the texture change of the scar area in different gray images, and determine the area and gray scale of the hyperplastic scar in the scar area when the hyperplastic scar appears, and determine the hyperplasia degree of the hyperplastic scar according to the area feature and the gray scale feature.

[0081] When the scar is analyzed, the hyperplasia of the scar also needs to be specifically analyzed. The hyperplasia of the scar is mainly the pathological scar caused by the imbalance of extracellular matrix metabolism, abnormal activation of fibroblasts and inflammatory reaction of the scar itself, and whether the scar hyperplasia appears needs to be analyzed at any time.

[0082] Further, in some embodiments of the present application, whether hyperplastic scar appears is determined according to the texture change of the scar area in different gray images, including: skeleton extraction is performed on the scar area in different gray images to determine the scar skeleton pixel points in each gray image; the scar skeleton pixel points belonging to the latter frame and not belonging to the former frame gray image in the adjacent two frames are taken as the hyperplasia pixel points; the ratio of the number of hyperplasia pixel points to the number of scar skeleton pixel points in the corresponding latter frame gray image is taken as the hyperplasia coefficient; when the hyperplasia coefficient is greater than a preset hyperplasia threshold, it is determined that hyperplastic scar appears, otherwise, it is determined that hyperplastic scar does not appear.

[0083] In the embodiment of the present application, the scar skeleton pixel points are determined by skeleton extraction. The skeleton extraction can specifically use morphological erosion to extract skeleton information, or a skeleton extraction algorithm can be directly used to realize skeleton analysis. In the embodiment of the present application, the scar hyperplasia generated in adjacent images is analyzed, so the scar skeleton pixel points that have in the latter frame and do not have in the former frame are taken as the hyperplasia pixel points.

[0084] The more the number of hyperplasia pixel points is, the greater the severity of the hyperplastic scar is. In the embodiment of the present application, the ratio of the number of hyperplasia pixel points to the number of scar skeleton pixel points in the corresponding latter frame gray image is taken as the hyperplasia coefficient.

[0085] The preset hyperplasia threshold is a threshold value of the hyperplasia coefficient, and in the embodiment of the present application, the preset hyperplasia threshold can be specifically, for example, 0.15, that is, when the hyperplasia coefficient is greater than 0.15, it is determined that the hypertrophic scar occurs, otherwise, it is determined that the hypertrophic scar does not occur, and the skeleton change can be an error caused when image analysis is performed.

[0086] When the hypertrophic scar occurs, the hyperplasia degree of the hypertrophic scar is determined according to the area feature and the gray scale feature, including: taking the hyperplasia coefficient as the area feature; calculating the gray scale average of all hyperplasia pixel points, and taking the reciprocal of the gray scale average as the gray scale feature index after normalization processing; and taking the product value of the hyperplasia coefficient and the gray scale feature index as the hyperplasia degree after normalization processing.

[0087] The greater the area of the hypertrophic scar area and the lower the gray scale value, the more serious the corresponding hypertrophic scar, and the lower the gray scale value, the darker the overall color of the hypertrophic scar, and the more serious the hypertrophic scar feature.

[0088] The hyperplasia degree is calculated based on this, and the greater the value of the hyperplasia degree, the more serious the hyperplasia effect.

[0089] The evaluation module 104 is used to adjust the recovery state index according to the hyperplasia degree in different gray scale images to obtain a recovery evaluation coefficient, and determine a recovery evaluation result according to the recovery evaluation coefficient at different sampling moments.

[0090] In the embodiment of the present application, the recovery state index can be adjusted, that is, the value is suppressed, and the greater the value of the hyperplasia degree, the worse the recovery state, at this time, the recovery state index needs to be reduced.

[0091] Further, in some embodiments of the present application, the product value of the hyperplasia degree and the preset influence weight is taken as an adjustment index, and the difference between the recovery state index and the adjustment index is taken as the recovery evaluation coefficient.

[0092] The preset influence weight is the influence weight of the hyperplasia degree on the recovery state index, and specifically, the preset influence weight can be, for example, 0.2, that is, the product of the hyperplasia degree and 0.2 is calculated as the adjustment index, and then the difference between the recovery state index and the adjustment index is taken as the recovery evaluation coefficient.

[0093] The recovery evaluation coefficient is an evaluation value obtained by combining the scar hyperplasia and the microvessel area texture feature, and the recovery evaluation coefficient can represent the overall state of the scar at the corresponding sampling moment, and the recovery evaluation of the scar can be realized based on the recovery evaluation coefficient.

[0094] According to the recovery evaluation coefficients at different sampling time points, the recovery evaluation result is determined, including: calculating the average value of the recovery evaluation coefficients at all sampling time points to obtain a total recovery index; when the value of the total recovery index is greater than 0.7, the recovery evaluation result is determined to be excellent, otherwise, the recovery evaluation result is determined to be general.

[0095] Of course, in other embodiments of the application, various other numerical division criteria can also be provided, or the total recovery index can be directly used as the recovery evaluation result, which is not limited.

[0096] The embodiment of the application determines the scar area and the microvessel area by acquiring the gray images of the dermatoscope at different sampling time points in the scar recovery stage, and then performs classification analysis, which is more accurate and reliable than directly analyzing the whole scar, so that the analysis result is more accurate; in the recovery analysis module, the blood vessel distribution characteristic index is determined according to the microvessel texture density and the distribution uniformity in the microvessel area, and the recovery state index is determined according to the change of the blood vessel distribution characteristic index in time sequence and the area change of the scar area, so that the recovery state index can be analyzed in two dimensions of microvessel texture characteristics and scar area characteristics, and the recovery state index is more accurate; in the hyperplasia analysis module, the hyperplasia degree of the hypertrophic scar is determined by analyzing the area and gray state of the hypertrophic scar; so that in the evaluation module, the recovery state index is adjusted according to the hyperplasia degree, and the recovery evaluation result is determined comprehensively. In summary, the application can evaluate the overall recovery state and the hyperplasia of the pathological scar area in time, effectively improving the accuracy and reliability of the evaluation result of the scar recovery.

[0097] It should be noted that: the above-mentioned embodiment of the application is only for description, and does not represent the advantages and disadvantages of the embodiment. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0098] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

Claims

1. A deep learning-based dynamic assessment system for postoperative scar recovery in plastic surgery, characterized in that, The system includes: The acquisition module is used to acquire grayscale images of dermatoscopes at different sampling times during the scar recovery stage, perform edge detection on the grayscale images to obtain edge images, and divide the scar region and the microvascular region around the scar from the edge images; The recovery analysis module is used to determine the vascular distribution characteristic index of the microvascular region based on the microvascular texture density and distribution uniformity in the microvascular region; and to determine the recovery status index at the sampling time based on the changes in the vascular distribution characteristic index in adjacent grayscale images in time sequence and the area change of the scar region. The hyperplasia analysis module is used to determine whether hyperplastic scars have appeared based on the texture changes of scar areas in different grayscale images. When hyperplastic scars appear, it determines the area and grayscale of the hyperplastic scar in the scar area, and determines the degree of hyperplasia of the hyperplastic scar based on the area and grayscale characteristics. The evaluation module is used to adjust the restoration status index according to the degree of proliferation in different grayscale images to obtain the restoration evaluation coefficient, and to determine the restoration evaluation result based on the restoration evaluation coefficient at different sampling times. Methods for determining the recovery state index at the sampling time include: The difference between the blood vessel distribution feature index in the grayscale image at any sampling time and the previous sampling time is normalized and used as the blood vessel recovery coefficient at the corresponding sampling time. The scar recovery coefficient is obtained by calculating the area difference between the previous sampling time and the corresponding sampling time in the time series and normalizing the result. The sum of the vascular recovery coefficient and the scar recovery coefficient is normalized and used as an indicator of recovery status. Methods for determining whether hypertrophic scars have occurred include: Skeleton extraction is performed on scar regions in images of different grayscale levels to determine the scar skeleton pixels in each grayscale image. Scar skeleton pixels that belong to the next frame but not to the previous frame in the grayscale image of two adjacent frames are taken as hyperplastic pixels. The ratio of the number of proliferating pixels to the number of scar skeleton pixels in the corresponding subsequent grayscale image is calculated as the proliferation coefficient. When the hyperplasia coefficient is greater than a preset hyperplasia threshold, hyperplastic scars are determined to have occurred; otherwise, hyperplastic scars are determined not to have occurred.

2. The deep learning-based dynamic assessment system for postoperative scar recovery in plastic surgery as described in claim 1, characterized in that, Edge detection is performed on the grayscale image to obtain an edge image, including: The grayscale image is subjected to edge detection based on the Canny edge detection algorithm to obtain edge pixels. The grayscale image is binarized based on the edge pixels to obtain an edge image, wherein the pixel value corresponding to the edge pixels is 1, and the pixel value corresponding to other pixels is 0.

3. The deep learning-based dynamic assessment system for postoperative scar recovery in plastic surgery as described in claim 1, characterized in that, The scar region and the microvascular region surrounding the scar are delineated from the edge image, including: Based on a pre-trained image feature recognition model, the scar area and the microvascular area surrounding the scar are identified.

4. The deep learning-based dynamic evaluation system for postoperative scar recovery in plastic surgery as described in claim 2, characterized in that, Based on the microvascular texture density and distribution uniformity in the microvascular region, the vascular distribution characteristic indicators of the microvascular region are determined, including: The proportion of edge pixels in the microvascular region is calculated as the microvascular texture density. The distribution uniformity index is determined based on the distance between any two edge pixels and the number distribution of edge pixels in different sub-regions of the microvascular region. The product of the microvascular texture density and distribution uniformity index is calculated and normalized to serve as the vascular distribution characteristic index of the microvascular region.

5. The deep learning-based dynamic evaluation system for postoperative scar recovery in plastic surgery as described in claim 4, characterized in that, Based on the distance between any two edge pixels and the distribution of edge pixels in different sub-regions within the microvascular region, a distribution uniformity index is determined, including: The microvascular region is divided into sub-regions of a preset size, and the variance of the number of edge pixels in each sub-region is determined as the dispersion coefficient of the number distribution. Calculate the mean distance and variance between any two edge pixels, and normalize the product of the mean distance and variance as the distance distribution dispersion coefficient. The sum of the dispersion coefficients of the quantity distribution and the dispersion coefficients of the distance distribution is used as the total dispersion index. The negative value of the total dispersion index is calculated and normalized to serve as the distribution uniformity index.

6. The deep learning-based dynamic evaluation system for postoperative scar recovery in plastic surgery as described in claim 1, characterized in that, The degree of hypertrophic scarring is determined based on area and grayscale characteristics, including: The growth coefficient is used as an area feature; the gray-scale mean of all grown pixels is calculated, and the negative of the gray-scale mean is normalized and used as a gray-scale feature index. The product of the growth coefficient and the gray-scale feature index is normalized and used as the degree of growth.

7. The deep learning-based dynamic assessment system for postoperative scar recovery in plastic surgery as described in claim 1, characterized in that, Based on the degree of proliferation in different grayscale images, the recovery status index is adjusted to obtain the recovery evaluation coefficient, including: The product of the degree of proliferation and the preset influence weight is used as the adjustment index; The difference between the recovery status index and the adjustment index is calculated and used as the recovery assessment coefficient.

8. The deep learning-based dynamic assessment system for postoperative scar recovery in plastic surgery as described in claim 1, characterized in that, Based on the recovery evaluation coefficients at different sampling times, the recovery evaluation results are determined, including: The mean of the recovery evaluation coefficients at all sampling times is calculated to obtain the overall recovery index; When the value of the total recovery index is greater than 0.7, the recovery assessment result is determined to be excellent; otherwise, the recovery assessment result is determined to be average.

Images