A skin disease three-dimensional imaging and lesion evolution trend simulation system

Abstract

The application relates to the technical field of image analysis, in particular to a skin disease three-dimensional imaging and lesion evolution trend simulation system. The system acquires state images of skin diseases; acquires a lesion area based on RGB channel values of pixel points in the state images; according to the dimensional change of the lesion area in each state image compared with that in a previous adjacent state image, the change degree of each dimension of the lesion area is acquired, and then the dominant weight of each dimension is determined to acquire a dominant dimension; according to the difference of the dominant dimensions of adjacent state images and the difference of the dominant weights of the dimensions, each evolution stage of the skin disease is determined; according to the influence of different filter kernel scales on the change degree of the dominant dimension in the evolution stage, the optimal filter kernel scale of each evolution stage is acquired to perform three-dimensional imaging and lesion evolution trend simulation of the skin disease. The application improves the accuracy of skin disease area recognition and lesion evolution trend simulation by adaptively selecting the optimal filter kernel scale of each evolution stage.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image analysis, in particular to a skin disease three-dimensional imaging and lesion evolution trend simulation system. BACKGROUND

[0002] Early diagnosis and dynamic monitoring of skin diseases rely on accurate analysis of lesion morphological evolution. Three-dimensional imaging technology provides comprehensive and detailed data support for observation and analysis of skin diseases. Three-dimensional imaging technology can be used for intelligent recognition and positioning of skin diseases and analysis of lesion morphological evolution trend. Three-dimensional imaging technology reconstructs a three-dimensional model of the skin surface through multi-angle image acquisition.

[0003] In actual situations, skin diseases have different evolution stages, and different evolution stages require different filter kernel scales to filter skin disease images. For example, a relatively large filter kernel scale is needed to smooth noise and avoid color interference in the color dominant evolution stage. A relatively small filter kernel scale is needed to filter skin disease images to retain boundary details in the shape dominant evolution stage. However, the prior art ignores the adjustment of the filter kernel scale for different evolution stages of skin diseases, and uses a fixed filter kernel scale to filter skin disease images, which cannot accurately represent the characteristic information of skin diseases, resulting in inaccurate construction of three-dimensional models of skin diseases, and further inaccurate simulation of lesion evolution trends of skin diseases. SUMMARY

[0004] To solve the technical problem of inaccurate three-dimensional model and lesion evolution trend simulation of skin diseases due to inaccurate filter kernel scale, the purpose of the present application is to provide a skin disease three-dimensional imaging and lesion evolution trend simulation system, and the technical solution adopted is as follows:

[0005] The present application provides a skin disease three-dimensional imaging and lesion evolution trend simulation system, which comprises the following steps:

[0006] An image acquisition module is used to acquire real-time state images of skin diseases, and acquire reference images of normal skin areas.

[0007] A lesion area acquisition module is used to filter out abnormal pixel points in each state image and determine the lesion area in each state image based on the RGB channel value difference between each pixel point in each state image and the pixel points in the reference image.

[0008] A dominant weight acquisition module is used to acquire the change degree of each dimension of each lesion area in each state image according to the change of each lesion area in each state image compared to each dimension in the previous adjacent state image, and to determine the dominant weight of each dimension in each state image.

[0009] an evolution stage acquisition module, configured to acquire a dominant dimension in each state image based on a dominant weight, and to filter segmented state images to determine each evolution stage of the skin disease according to a difference in the dominant dimension between each state image and its previous adjacent state image and a difference in the dominant weight of each dimension;

[0010] an optimal filter kernel scale acquisition module, configured to acquire an optimal filter kernel scale for each evolution stage according to an influence of different filter kernel scales on a variation degree of the dominant dimension of each lesion region in each state image in each evolution stage;

[0011] a processing module, configured to perform three-dimensional imaging and lesion evolution trend simulation of the skin disease based on the optimal filter kernel scale.

[0012] Further, the method for acquiring the abnormal pixel points comprises the following steps:

[0013] for any one of the RGB channels, taking an average of the channel value of all pixel points in the reference image as an overall reference value of the channel;

[0014] normalizing a difference between the channel value of each pixel point in the reference image and the overall reference value to obtain a tolerance degree of the channel of each pixel point in the reference image;

[0015] adding the maximum tolerance degree of each RGB channel of all pixel points in the reference image to obtain a color overall tolerance degree;

[0016] for any pixel point in any state image, normalizing a difference between an addition result of the tolerance degrees of each RGB channel of the pixel point and the color overall tolerance degree to obtain an abnormality degree of the pixel point;

[0017] when the abnormality degree is greater than a preset abnormality degree threshold, the corresponding pixel point is an abnormal pixel point.

[0018] Further, the method for acquiring the variation degree comprises the following steps:

[0019] for any lesion region in any state image and any dimension of the lesion region, taking a previous adjacent state image of the state image as a target image, and taking a difference between data corresponding to the dimension of the lesion region in the state image and the target image as a first difference;

[0020] normalizing a ratio of the first difference to data corresponding to the dimension of the lesion region in the target image to obtain a variation degree of the dimension of the lesion region in the state image.

[0021] Further, the method for acquiring the dominant weight comprises the following steps:

[0022] For any state image, a covariance matrix corresponding to the degree of change of all dimensions of all lesion regions in the state image is obtained by a principal component analysis algorithm, and a feature vector corresponding to a maximum eigenvalue of the covariance matrix is taken as a principal component vector;

[0023] A result of normalizing an absolute value of each element in the principal component vector is taken as a dominant weight of the corresponding dimension in the state image.

[0024] Further, the method for obtaining the segmented state image is:

[0025] For any state image, a dominant weight difference of a dominant dimension between the state image and a previous adjacent state image is obtained as a first evolution analysis value of the state image; wherein, when the dominant dimension of the state image and the previous adjacent state image is different, the first evolution analysis value is defaulted as 1;

[0026] A mean value of the dominant weight difference of each dimension between the state image and the previous adjacent state image is obtained as a second evolution analysis value of the state image;

[0027] A result of normalizing a product of the first evolution analysis value and the second evolution analysis value is taken as an evolution degree of the state image;

[0028] When the evolution degree is greater than a predicted evolution degree threshold value, the corresponding state image is taken as a segmented state image.

[0029] Further, the method for obtaining the evolution stage is:

[0030] A time corresponding to the first state image is taken as a starting time, and a time corresponding to the last state image is taken as a terminal time; a time period constituted by the starting time and the terminal time is taken as an overall stage;

[0031] A previous adjacent time of a time corresponding to each segmented state image is taken as a truncation time to divide the overall stage, and each local stage after the division is taken as each evolution stage.

[0032] Further, the method for obtaining the optimal filter kernel scale is:

[0033] For any evolution stage, a dominant dimension appearing most frequently in the dominant dimension in all segmented state images in the evolution stage is taken as a dominant dimension in all state images in the evolution stage;

[0034] For any filter kernel scale, each state image in the evolution stage is filtered by the filter kernel scale to obtain a filtered state image;

[0035] Obtaining the mean value of the change degree of the dominant dimension of all the lesion regions in all the filtered state images as the filtering degree of the filtering kernel scale to the evolution stage;

[0036] Taking the filtering kernel scale corresponding to the maximum filtering degree as the optimal filtering kernel scale of the evolution stage.

[0037] Further, the method for obtaining the dominant dimension is:

[0038] Taking the dimension corresponding to the maximum dominant weight in each state image as the dominant dimension in each state image.

[0039] Further, the method for obtaining the lesion region is:

[0040] For any state image, the connected domain formed by the abnormal pixel points in the state image is taken as an abnormal region.

[0041] For any abnormal region, the region corresponding to the abnormal region after the first opening operation and then the closing operation is taken as a lesion region.

[0042] Further, the method for obtaining the state image is:

[0043] For any time, the image obtained by registering and splicing the skin disease images obtained from multiple angles at the time is taken as the state image at the time.

[0044] The present application has the following beneficial effects:

[0045] The application firstly screens out the abnormal pixel points in each state image based on the RGB channel value difference between each pixel point in each state image and the pixel points in the reference image, prepares for subsequent determination of the lesion area of the skin disease in each state image, and then determines the lesion area in each state image, improves the efficiency of subsequent analysis of the change of the lesion area, and is beneficial to subsequent accurate division of each evolution stage of the skin disease; in order to accurately analyze the change of each lesion area, and then according to the change of each lesion area in each state image compared with each dimension in the previous adjacent state image, the change degree of each dimension of each lesion area in each state image is obtained, the change of each dimension of each lesion area is accurately reflected, in order to accurately divide each evolution stage of the skin disease, and then the dominant weight of each dimension in each state image is determined through the change degree of each dimension of each lesion area in each state image, and the overall change of each dimension in each state image is accurately reflected; then the dominant dimension in each state image is accurately obtained based on the dominant weight, the main changing dimension in each state image is accurately reflected, and the preparation for subsequent determination of each evolution stage of the skin disease is prepared; then, according to the difference between the dominant dimensions of each state image and its previous adjacent state image, and the difference of the dominant weight of each dimension, the segmented state image is accurately screened out, the segmentation image between different evolution stages is accurately determined, and then the evolution stage of the skin disease is determined; further, according to the influence of different filter kernel scales on the change degree of the dominant dimension of each lesion area in each state image in each evolution stage, the optimal filter kernel scale of each evolution stage is obtained, which is beneficial to accurate filtering processing of the skin disease image in the evolution stage, and then the three-dimensional imaging and lesion evolution trend simulation of the skin disease are carried out based on the optimal filter kernel scale, which effectively improves the accuracy of the three-dimensional model construction of the skin disease, accurately simulates the lesion evolution trend of the skin disease, and is beneficial to accurate analysis of the condition of the skin disease. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, hereinafter, a brief introduction will be given to 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 for those skilled in the art, other drawings can also be obtained without creative labor.

[0047] Figure 1 The structural block diagram of a skin disease three-dimensional imaging and lesion evolution trend simulation system provided by an embodiment of the present application;

[0048] Figure 2 The flow chart of a lesion area acquisition method provided by an embodiment of the present application;

[0049] Figure 3 A schematic diagram of a computer device provided by one embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined object, the following describes in detail the specific implementation, structure, features and effects of a skin disease three-dimensional imaging and lesion evolution trend simulation system according to the present application, with reference to the accompanying drawings and preferred embodiments. 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 specific scheme of the skin disease three-dimensional imaging and lesion evolution trend simulation system provided by the present application is described in detail below with reference to the accompanying drawings.

[0053] Embodiment 1:

[0054] The present application provides a skin disease three-dimensional imaging and lesion evolution trend simulation system, please refer to Figure 1 , which shows the structure block diagram of a skin disease three-dimensional imaging and lesion evolution trend simulation system provided by one embodiment of the present application, the system includes: image acquisition module 10, lesion area acquisition module 20, dominant weight acquisition module 30, evolution stage acquisition module 40, optimal filter kernel scale acquisition module 50 and processing module 60.

[0055] The image acquisition module 10 is used to acquire the state image of the skin disease in real time; and acquire the reference image of the normal area of the skin.

[0056] Specifically, the present embodiment takes a skin disease patient as an example for analysis, and the subsequent analysis is also based on the skin disease of the skin disease patient. In order to accurately analyze the lesion evolution trend of the skin disease, the present embodiment acquires skin disease images from multiple angles at each time point, and then acquires the image obtained by registering and splicing the skin disease images acquired from multiple angles at each time point as the state image of the skin disease at each time point. The registration and splicing of the skin disease images acquired from multiple angles at each time point are known technologies and will not be described herein. The present embodiment sets the time interval between adjacent two time points as 12 hours, and the implementer can set the time interval between adjacent two time points according to the actual situation, which is not limited herein. The angles at which the skin disease images are acquired at each time point can be set according to the actual situation, which is not limited herein, but it is necessary to ensure that the skin disease images collected at each time point cover the entire skin disease area.

[0057] It is known that the distribution area of the skin disease is irregular in actual situation. In order to accurately determine the lesion area in each state image, the present embodiment acquires the image of the normal skin area of the above-mentioned skin disease patient as a reference image. Because the skin color of the same person is similar, and the skin color of different persons is different, the lesion area in each state image can be accurately acquired by comparing each state image with the reference image, which is beneficial to more efficient analysis of the change of the skin disease in the subsequent.

[0058] The lesion area acquisition module 20 is configured to filter out abnormal pixel points in each state image based on the difference between the RGB channel values of each pixel point in each state image and the pixel points in the reference image, and determine the lesion area in each state image.

[0059] Specifically, it is known that the lesion area of the skin disease will appear red and swollen, damaged, etc., and there is a significant color difference with the normal skin area. Therefore, the abnormal pixel points in each state image can be filtered out and the lesion area in each state image can be determined by the difference between the color features of each pixel point in each state image and the color features of the pixel points in the reference image. Therefore, the present embodiment filters out the abnormal pixel points in each state image based on the difference between the RGB channel values of each pixel point in each state image and the pixel points in the reference image, and determines the lesion area in each state image.

[0060] Preferably, in one implementation manner of the present embodiment, the method for acquiring the lesion area can refer to Figure 2 which shows a method flowchart for acquiring the lesion area provided by the present embodiment, and the method comprises the following steps:

[0061] Step S201: Acquiring abnormal pixel points.

[0062] In view of the fact that the skin also has certain color unevenness in a healthy state, the embodiment first acquires the maximum fluctuation degree of the skin color in the reference image as a reference of the skin in a healthy state, for subsequent comparison with the color fluctuation degree of each pixel point in each state image, so as to accurately screen out abnormal pixel points with obvious color fluctuation in each state image, which is beneficial to subsequent accurate acquisition of the lesion region in each state image.

[0063] In an implementable manner of the embodiment, the abnormal pixel point acquisition method is as follows: for any channel in the RGB channel, the mean value of the channel value of all pixel points in the reference image is taken as the overall reference value of the channel; the result of normalizing the absolute value of the difference between the channel value of each pixel point in the reference image and the overall reference value is taken as the tolerance degree of the channel of each pixel point in the reference image; the embodiment normalizes the above absolute value by using the norm normalization function. The sum of the maximum tolerance degrees of each RGB channel of all pixel points in the reference image is taken as the overall color tolerance degree, that is, the maximum fluctuation degree of the color of the skin in a healthy state;

[0064] For any pixel point in any state image, the result of normalizing the difference between the sum of the tolerance degrees of each RGB channel of the pixel point and the overall color tolerance degree is taken as the abnormality degree of the pixel point; the embodiment normalizes the above difference by using the norm normalization function. The greater the abnormality degree, the greater the color fluctuation of the pixel point compared with the color fluctuation of the normal skin, and the more likely the pixel point is an abnormal pixel point; furthermore, the embodiment sets the preset abnormality degree threshold value to 0.6, and the implementer can set the size of the preset abnormality degree threshold value according to the actual situation, which is not limited herein. When the abnormality degree is greater than the preset abnormality degree threshold value, the corresponding pixel point is an abnormal pixel point.

[0065] At this point, the abnormal pixel points in each state image are acquired.

[0066] Step S202: Acquire the lesion region.

[0067] After determining the abnormal pixel points in each state image, the connected domain formed by the abnormal pixel points in each state image is taken as an abnormal region; wherein, the abnormal region is essentially a lesion region, but considering that in actual situations, part of the noise pixel points may be mistakenly regarded as abnormal pixel points, and in order to accurately acquire the lesion region in each state image, the embodiment first performs an opening operation on each abnormal region to improve the purity of the lesion region and reduce false detection; after the opening operation, in order to ensure the integrity of the lesion region, a closing operation is further performed, and finally the region corresponding to each abnormal region after the opening operation and the closing operation is taken as the lesion region, effectively ensuring the integrity and accuracy of the lesion region.

[0068] Thus, the lesion area in each state image is obtained.

[0069] The dominant weight obtaining module 30 is configured to obtain the change degree of each dimension of each lesion area in each state image according to the change of each lesion area in each state image compared with each dimension of the previous adjacent state image, and further determine the dominant weight of each dimension in each state image.

[0070] Specifically, the lesion area of the skin disease changes in many aspects, such as color, area and shape, etc. in the evolution process. In order to analyze the change of each lesion area, the color, area and shape of each lesion area are taken as the dimension of each lesion area in the embodiment, and the implementer can set the dimension of the lesion area according to the actual situation, which is not limited herein. The number and type of the dimension of each lesion area are certainly the same. In order to analyze the change of each dimension of each lesion area in each state image, so as to determine the evolution stage of the skin disease, the change degree of each dimension of each lesion area in each state image is obtained according to the change of each lesion area in each state image compared with each dimension of the previous adjacent state image in the embodiment, which accurately reflects the change of each dimension of each lesion area in each state image, and prepares for obtaining the overall change of each dimension in each state image. Further, the dominant weight of each dimension in each state image is determined by the change degree of each dimension of each lesion area in each state image in the embodiment, and the greater the dominant weight is, the greater the change of the corresponding dimension in the corresponding state image is, which is beneficial to accurately determine the main change dimension in each state image, so as to accurately obtain the evolution stage of the skin disease.

[0071] Preferably, in one implementation manner of the embodiment, the change degree obtaining method is that: for any lesion area and any dimension of the lesion area in any state image, the previous adjacent state image of the state image is taken as a target image, the absolute value of the difference between the data corresponding to the dimension of the lesion area in the state image and the target image is taken as a first difference, and the result of normalizing the ratio of the first difference to the data corresponding to the dimension of the lesion area in the target image is taken as the change degree of the dimension of the lesion area in the state image. The above ratio is normalized by the norm normalization function in the embodiment.

[0072] It should be noted that the data corresponding to the color dimension of the lesion region in the state image is set as the mean value of the abnormality degree of all pixel points of the lesion region in the state image. Since the value range of the abnormality degree is 0 to 1, it can be inferred that the data corresponding to the color dimension of the lesion region in the target image is 0, that is, the denominator is 0. In order to avoid the case that the denominator is 0, in the embodiment, the addition result of the data corresponding to the color dimension of the lesion region in the target image and a preset constant is taken as the denominator, wherein the preset constant is greater than 0, and the embodiment sets the preset constant to 1. The implementer can set the size of the preset constant according to the actual situation, which is not limited herein. The data corresponding to the area dimension of the lesion region in the state image is set as the number of all pixel points in the lesion region in the state image. The data corresponding to the shape dimension of the lesion region in the state image is set as the addition result of the Euclidean distance between all edge pixel points on the outer edge line of the lesion region in the state image and the centroid of the lesion region in the state image. The centroid and the Euclidean distance are obtained by known technologies, and will not be described herein.

[0073] In addition, the position of the lesion region in the state image and the target image is obtained by the stream method. The stream method is a known technology, and will not be described herein. If the lesion region is a new region in the state image, that is, does not exist in the target image, the normalization result of the data corresponding to each dimension of the lesion region in the state image is taken as the change degree of each dimension of the lesion region in the state image. The data corresponding to each dimension of the lesion region in the state image is normalized by the norm normalization function. If the lesion region in the state image is the overall region after expansion of multiple lesion regions in the target image, the mean value of the abnormality degree of all pixel points of the corresponding lesion region in the target image is taken as the data corresponding to the color dimension of the lesion region in the target image. The number of all pixel points in the corresponding lesion region in the target image is taken as the data corresponding to the area dimension of the lesion region in the target image. The mean value of the addition result of the Euclidean distance between all edge pixel points on the outer edge line of the corresponding lesion region in the target image and the centroid of the lesion region is taken as the data corresponding to the shape dimension of the lesion region in the target image.

[0074] At this point, the change degree of each dimension of each lesion region in each state image is obtained. It should be noted that since the first state image does not have a previous adjacent state image, the change degree of each dimension of each lesion region in the first state image is not obtained.

[0075] Preferably, in one implementation manner of the present embodiment, the method for obtaining the dominant weight is that, for any state image, a covariance matrix corresponding to the variation degree of all dimensions of all lesion regions in the state image is obtained by a principal component analysis algorithm, and a feature vector corresponding to the largest eigenvalue of the covariance matrix is taken as a principal component vector; wherein the principal component analysis algorithm, the eigenvalue and the feature vector are all known technologies, and will not be described in detail. Each element in the principal component vector corresponds to a unique dimension, and then the result of normalizing the absolute value of each element in the principal component vector is taken as the dominant weight of the corresponding dimension in the state image. The present embodiment takes the ratio of the absolute value of each element in the principal component vector to the sum of the absolute values of all elements in the principal component vector as the result of normalizing the absolute value of each element in the principal component vector.

[0076] Thus, the dominant weight of each dimension in each state image is obtained.

[0077] The evolution stage obtaining module 40 is configured to obtain the dominant dimension in each state image based on the dominant weight, and determine each evolution stage of the skin disease according to the difference between the dominant dimension of each state image and its previous adjacent state image, and the difference of the dominant weight of each dimension.

[0078] Specifically, the main change dimension of the lesion in different evolution stages of the skin disease is different, for example, in a certain evolution stage, the area of the lesion region may change obviously, or the color of the lesion region may change obviously. It is known that in order to more accurately determine the lesion region and accurately simulate the evolution of the lesion region, the filter kernel scale corresponding to the skin disease image in different evolution stages should be self-adaptive according to the actual situation, for example, when the main change dimension in a certain evolution stage is the color dimension, if the filter kernel scale is too small, too many colors will be caused, which will affect the accurate filtering of the lesion region. Therefore, it is necessary to obtain each evolution stage of the skin disease. Considering that there are obvious differences in the main change dimension between different evolution stages, and there are also obvious differences between the same dimensions, therefore, the present embodiment first obtains the dominant dimension, i.e. the main change dimension, in each state image based on the dominant weight, and then screens out the segmented state image according to the difference between the dominant dimension of each state image and its previous adjacent state image, and the difference of the dominant weight of each dimension, and further determines each evolution stage of the skin disease.

[0079] Preferably, in an implementable manner of the embodiment, the method for obtaining the dominant dimension is: taking the dimension corresponding to the maximum dominant weight in each state image as the dominant dimension in the state image. If the maximum dominant weight in a state image corresponds to at least two dimensions, then optionally taking one of the dimensions as the dominant dimension in the state image.

[0080] Preferably, in an implementable manner of the embodiment, the method for obtaining the segmented state image is: for any state image, obtaining the absolute value of the difference between the dominant weight of the dominant dimension of the state image and the dominant weight of the dominant dimension of the previous adjacent state image as the first evolution analysis value of the state image; wherein, when the dominant dimensions of the state image and the previous adjacent state image are different, the first evolution analysis value is defaulted to 1; obtaining the mean value of the absolute values of the differences between the dominant weights of each dimension of the state image and the previous adjacent state image as the second evolution analysis value of the state image; when the first evolution analysis value and the second evolution analysis value are both larger, it indicates that the state image is more likely to be a segmented image of different evolution stages, and then the embodiment takes the normalized result of the product of the first evolution analysis value and the second evolution analysis value as the evolution degree of the state image; the embodiment normalizes the product of the first evolution analysis value and the second evolution analysis value by the norm normalization function. The greater the evolution degree, the more likely the state image is a segmented image of different evolution stages, and then the embodiment sets the prediction evolution degree threshold to 0.7, and the implementer can set the size of the prediction evolution degree threshold according to the actual situation, which is not limited here. When the evolution degree is greater than the prediction evolution degree threshold, the corresponding state image is taken as the segmented state image.

[0081] At this point, all segmented state images up to the current time are obtained, which is conducive to accurately determining the evolution stages of the skin disease.

[0082] Preferably, in an implementable manner of the embodiment, the method for obtaining the evolution stage is: taking the time corresponding to the first state image as the starting time, and taking the time corresponding to the last state image as the ending time; taking the time period constituted by the starting time and the ending time as the overall stage; taking the previous adjacent time of the time corresponding to each segmented state image as the truncation time to divide the overall stage, and taking each local stage after the division as each evolution stage.

[0083] At this point, the evolution stages of the skin disease are obtained.

[0084] The optimal filter kernel scale obtaining module 50 is configured to obtain the optimal filter kernel scale of each evolution stage according to the influence of different filter kernel scales on the change degree of the dominant dimension of each lesion area in each state image in each evolution stage.

[0085] Specifically, the changes of the lesion regions in the state images in the same evolution stage are similar, and therefore, the embodiment obtains the optimal filter kernel scale of each evolution stage according to the influence of different filter kernel scales on the change degree of the dominant dimension of each lesion region in each state image in each evolution stage, so that the three-dimensional lesion region of the skin disease is accurately constructed through the three-dimensional imaging technology, and the lesion evolution trend is accurately simulated.

[0086] Preferably, in an implementable manner of the embodiment, the method for obtaining the optimal filter kernel scale is as follows: for any evolution stage, the dominant dimension with the highest occurrence frequency in the dominant dimension of all the segmented state images in the evolution stage is taken as the dominant dimension of all the state images in the evolution stage. For any filter kernel scale, the state image in the evolution stage is filtered through the filter kernel scale to obtain a filtered state image; the mean value of the change degree of the dominant dimension of all the lesion regions in all the filtered state images is taken as the filtering degree of the filter kernel scale for the evolution stage; the greater the filtering degree, the more likely the filter kernel scale is the optimal filter kernel scale for the evolution stage, because a filter kernel scale that is too large or too small will cause the change degree of the dominant dimension of the lesion region to deviate too small, for example, if the dominant dimension is the shape dimension, a filter kernel scale that is too large will blur the edge details, and a filter kernel scale that is too small will retain noise, resulting in insignificant shape changes. Therefore, the filter kernel scale corresponding to the maximum filtering degree is taken as the optimal filter kernel scale for the evolution stage. It should be noted that the filter kernel scales of the embodiment are set to 3, 5, 7, 9, and 11, and the implementer can set the filter kernel scales according to the actual situation, which are not limited herein. and

[0087] Thus, the optimal filter kernel scale of each evolution stage is obtained.

[0088] The processing module 60 is configured to perform three-dimensional imaging of the skin disease and simulate the lesion evolution trend based on the optimal filter kernel scale.

[0089] Specifically, each skin disease image at each time in each evolution stage is filtered through the optimal filter kernel scale of each evolution stage, and the filtered skin disease image at each time is processed through the three-dimensional imaging technology to construct the three-dimensional imaging of the skin disease at each time, and the lesion evolution trend of the skin disease is accurately simulated, which is conducive to accurately analyzing the condition of the skin disease.

[0090] ​To sum up, the embodiment obtains a state image of a skin disease; obtains a lesion area based on an RGB channel value of a pixel point in the state image; obtains a change degree of each dimension of the lesion area according to a change of the lesion area in each state image compared with a dimension in a previous adjacent state image, and then determines a dominant dimension by determining a dominant weight of each dimension; determines each evolution stage of the skin disease according to a difference of the dominant dimensions of adjacent state images and a difference of the dominant weights of the dimensions; and obtains an optimal filter kernel scale of each evolution stage according to an influence of different filter kernel scales on the change degree of the dominant dimension in the evolution stage, to perform three-dimensional imaging and lesion evolution trend simulation of the skin disease. The optimal filter kernel scale of each evolution stage is adaptively obtained, so that the accuracy of skin disease area recognition and lesion evolution trend simulation is improved.

[0091] Embodiment 2

[0092] The application further provides a skin disease three-dimensional imaging and lesion evolution trend simulation device, which comprises a memory and a processor, wherein the memory stores executable program codes, and the processor is used for calling and executing the executable program codes to execute the skin disease three-dimensional imaging and lesion evolution trend simulation system provided in the embodiments. The device can be a chip, an assembly or a module. The chip can comprise a processor and a memory connected to each other. The memory is used for storing instructions, and when the processor calls and executes the instructions, the chip can execute the skin disease three-dimensional imaging and lesion evolution trend simulation system provided in the embodiments.

[0093] In addition, the embodiments of the application also protect a computer device, please refer to Figure 3 The computer device comprises a memory 401, a processor 402 and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device can execute any one of the skin disease three-dimensional imaging and lesion evolution trend simulation systems introduced above.

[0094] Embodiment 3

[0095] The application further provides a computer readable storage medium, which stores computer program codes. When the computer program codes run on a computer, the computer executes the related method steps to realize the skin disease three-dimensional imaging and lesion evolution trend simulation system provided in the embodiments.

[0096] Embodiment 4

[0097] The application further provides a computer program product, which, when running on a computer, enables the computer to execute the above related steps to realize the skin disease three-dimensional imaging and lesion evolution trend simulation system provided by the above embodiment.

[0098] Among them, the device, computer readable storage medium, computer program product or chip provided by the embodiment are used to execute the corresponding method provided above, so the beneficial effects that can be achieved are referred to the beneficial effects in the corresponding method provided above, which will not be described here.

[0099] It should be noted that the above-mentioned sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. 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 can be advantageous.

[0100] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the difference from other embodiments.

Claims

1. A three-dimensional imaging and lesion evolution trend simulation system for skin diseases, characterized in that, The system includes the following steps: The image acquisition module is used to acquire real-time images of skin diseases and to acquire reference images of normal skin areas. The lesion area acquisition module is used to filter out abnormal pixels in each status image based on the difference in RGB channel values ​​between each pixel in each status image and the pixel in the reference image, thereby determining the lesion area in each status image. The dominant weight acquisition module is used to obtain the degree of change of each dimension of each lesion region in each state image based on the changes of each lesion region in each state image compared with its changes in each dimension in the previous adjacent state image, and then determine the dominant weight of each dimension in each state image. The evolution stage acquisition module is used to acquire the dominant dimension in each state image based on the dominant weight. Based on the difference in dominant dimension between each state image and its previous adjacent state image, as well as the difference in dominant weight of each dimension, the segmented state images are filtered to determine the various evolution stages of the skin disease. The optimal filter kernel scale acquisition module is used to obtain the optimal filter kernel scale for each evolution stage based on the influence of different filter kernel scales on the degree of change of the dominant dimension of each lesion region in each state image within each evolution stage. The processing module is used for three-dimensional imaging of skin diseases and simulation of lesion evolution trends based on the optimal filter kernel scale; The method for obtaining the dominant weight is as follows: For any state image, the covariance matrix corresponding to the degree of change of all dimensions of all lesion regions in the state image is obtained by the principal component analysis algorithm, and the eigenvector corresponding to the largest eigenvalue of the covariance matrix is ​​used as the principal component vector. The result of normalizing the absolute value of each element in the principal component vector is used as the dominant weight of the corresponding dimension in the state image. The method for obtaining the optimal filter kernel scale is as follows: For any evolution stage, the dominant dimension that appears most frequently in all segmented state images within that evolution stage is taken as the dominant dimension in all state images within that evolution stage. For any given filter kernel scale, each state image in the evolution stage is filtered using that filter kernel scale to obtain a filtered state image. The mean value of the degree of change of the dominant dimension of all lesion regions in all the filtered state images is obtained as the filtering degree of the filtering kernel scale for this evolution stage; The filter kernel size corresponding to the maximum filtering degree is taken as the optimal filter kernel size for this evolution stage.

2. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the abnormal pixels is as follows: For any channel in the RGB channels, the average value of that channel for all pixels in the reference image is used as the overall reference value for that channel. The result of normalizing the difference between the channel value of each pixel in the reference image and the overall reference value is used as the tolerance level of the channel of each pixel in the reference image. The sum of the maximum tolerance levels for each RGB channel of all pixels in the reference image is used as the overall color tolerance level. For any pixel in any state image, the sum of the tolerance levels of each RGB channel of that pixel and the difference between the sum and the overall color tolerance level are normalized to determine the degree of anomaly of that pixel. When the degree of abnormality exceeds the preset abnormality threshold, the corresponding pixel is considered an abnormal pixel.

3. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the degree of change is as follows: For any lesion region and any dimension of the lesion region in any state image, the previous adjacent state image of the state image is taken as the target image, and the difference between the data corresponding to that dimension of the lesion region in the state image and the target image is taken as the first difference. The result of normalizing the first difference with the ratio of the corresponding dimension of the lesion region in the target image is taken as the degree of change of that dimension of the lesion region in the state image.

4. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the segmented state image is as follows: For any state image, the difference in dominant weights of the dominant dimension between the state image and its previous neighboring state image is obtained as the first evolution analysis value of the state image; wherein, when the dominant dimension of the state image is different from that of its previous neighboring state image, the first evolution analysis value is 1 by default. The mean of the difference in dominant weights for each dimension between the current state image and its previous neighboring state image is obtained as the second evolution analysis value of the current state image. The normalized product of the first evolution analysis value and the second evolution analysis value is taken as the degree of evolution of the state image. When the degree of evolution exceeds the predicted degree of evolution threshold, the corresponding state image is used as the segmented state image.

5. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the evolution stage is as follows: The time corresponding to the first state image is taken as the start time, and the time corresponding to the last state image is taken as the end time; the time period consisting of the start time and the end time is taken as the overall stage. The overall stage is divided by taking the time immediately preceding the time corresponding to each segmented state image as the cutoff time, and each local stage after the division is taken as each evolution stage.

6. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the dominant dimension is as follows: The dimension corresponding to the largest dominant weight in each state image is taken as the dominant dimension in each state image.

7. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the lesion area is as follows: For any state image, the connected components formed by the abnormal pixels in the state image are all regarded as abnormal regions. For any abnormal region, the region corresponding to the abnormal region after performing an opening operation followed by a closing operation is taken as the lesion region.

8. The three-dimensional imaging and lesion evolution trend simulation system for skin diseases as described in claim 1, characterized in that, The method for obtaining the state image is as follows: For any given moment, the image obtained by registering and stitching together skin disease images acquired from multiple angles at that moment is taken as the state image at that moment.

Images