Method for evaluating curative effect of smearing medicine used by skin disease patient

Abstract

The invention provides a method for evaluating the curative effect of applying a drug to a skin disease patient, belongs to the technical field of skin images, and aims at obtaining a high-quality skin image through a standardized skin imaging system and extracting an interested skin lesion area by adopting a segmentation algorithm based on deep learning. Then illumination normalization processing is carried out on the lesion area, a color representation center in a CIE L * a * b * color space is calculated, the color representation center is mapped to a standard camera obscura environment, and a corrected color center is obtained; and then comparing the corrected color center with color standard values of various lesion degrees established in advance to determine the current lesion degree. And finally, calculating the change rate of the lesion degree index according to the characteristic indexes of the lesion area, including the area proportion, the color difference and the lesion degree index, so as to determine the curative effect grade. According to the method, the technical problem that the accuracy of evaluation of the curative effect of the smeared medicine is influenced as the existing skin image acquisition is often influenced by the environment illumination condition is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of skin images, and in particular, relates to a method for evaluating the efficacy of topical medication used by patients with skin diseases. Background Art

[0002] Skin disease is a common clinical disease, and many types of skin diseases require continuous medication for treatment. The evaluation of drug efficacy is the key to clinical management of skin disease patients and directly affects the subsequent treatment strategy. Traditional evaluation of skin disease efficacy mainly relies on the subjective experience of doctors, and is judged by visually observing the skin lesion area, color and other characteristics. It has problems such as strong subjectivity, poor repeatability and low efficiency. In addition, since the symptoms of many skin diseases are highly dynamic, observation at a single time point cannot fully reflect the changes in the disease, and it is difficult to accurately evaluate the efficacy of drug treatment. In response to the above problems, the prior art has proposed some objective evaluation methods based on image analysis. These methods quantitatively analyze indicators such as skin lesion area and color by segmenting and extracting skin images, thereby obtaining some numerical evaluation results. However, these methods generally have a core problem, that is, the influence of light. Since the acquisition of skin images is often affected by ambient light conditions, the color characteristics of skin images acquired at different time points and different shooting angles may vary greatly, which will seriously affect the subsequent quantitative analysis results and reduce the accuracy and reliability of the evaluation method. Summary of the invention

[0003] In view of this, the present invention provides a method for evaluating the efficacy of smeared medications used by patients with skin diseases, which can solve the technical problem that the acquisition of existing skin images is often affected by ambient lighting conditions, thereby affecting the accuracy of the evaluation of the efficacy of smeared medications.

[0004] The present invention is achieved in that:

[0005] A first aspect of the present invention provides a method for evaluating the efficacy of topical medication used by patients with skin diseases, comprising the following steps:

[0006] S10, using a standardized skin imaging system to obtain a high-resolution digital image of the skin area to be evaluated, and preprocessing the digital image to obtain a first image;

[0007] S20, performing semantic segmentation on the first image using a deep learning-based image segmentation algorithm to extract a skin lesion area of ​​interest;

[0008] S30, performing illumination normalization processing on the skin lesion region of interest to obtain an illumination-invariant skin surface image;

[0009] S40, calculating the color representation center of the skin lesion degree in the CIE L*a*b* color space according to the skin appearance image with unchanged illumination;

[0010] S50, mapping the color representation center of the skin lesion degree in a dark box environment under a standard white light source to obtain a corrected color center;

[0011] S60, obtaining the standard value of the color representation center of each skin lesion degree in the dark box environment, recorded as the lesion degree color standard value; calculating the proximity probability between the corrected color center and each lesion degree color standard value, and taking the lesion degree corresponding to the lesion degree color standard value with the highest proximity probability as the current lesion degree of the skin area to be evaluated;

[0012] S70, calculating characteristic indicators of the skin lesion area, including the area ratio, lesion color difference, and lesion severity index;

[0013] S80. Calculate the change rate of the lesion severity index according to the characteristic index of the skin lesion area, and determine the efficacy grade according to the change rate.

[0014] On the basis of the above technical solution, the method for evaluating the efficacy of smear medication for patients with skin diseases of the present invention can also be improved as follows:

[0015] The preprocessing includes color correction, denoising and geometric correction.

[0016] Furthermore, the deep learning-based image segmentation algorithm adopts U-Net or Mask R-CNN.

[0017] The specific steps of step S10 include:

[0018] Step 101: Use a standardized skin imaging system to obtain a high-resolution digital image of the skin area to be evaluated. The imaging system usually includes a professional digital camera, a light source, and a fixed shooting platform to ensure that each image collected has consistent lighting conditions and imaging parameters.

[0019] Step 102: preprocess the acquired original digital image. Preprocessing includes steps such as color correction, denoising and geometric correction. Color correction aims to eliminate color distortion caused by factors such as uneven lighting or camera color deviation. Denoising uses a common low-pass filtering algorithm to eliminate noise components in the image. The purpose of geometric correction is to correct image geometric distortion caused by factors such as camera tilt or shooting angle.

[0020] Step 103: After these preprocessing steps, a high-quality, interference-free digital image of the skin area can be obtained, laying the foundation for subsequent computational analysis.

[0021] The specific steps of step S20 include:

[0022] Step 201: Perform semantic segmentation on the preprocessed first image using an image segmentation algorithm based on deep learning. It is recommended to use common segmentation network models such as U-Net or Mask R-CNN.

[0023] Step 202: Through these deep learning-based segmentation algorithms, the skin lesion area of ​​interest can be accurately extracted from the overall image. A large number of labeled skin images need to be used for training in advance to improve the recognition accuracy of the lesion area.

[0024] Step 203: After this step, an accurate mask image of the skin lesion area can be obtained.

[0025] The specific steps of step S30 include:

[0026] Step 301: Perform illumination normalization processing on the extracted skin lesion area, where homomorphic filtering technology and color constancy algorithm can be used.

[0027] Step 302: Homomorphic filtering can effectively suppress the non-uniform illumination components in the image, thereby obtaining an apparent skin image with constant illumination. The color constancy algorithm can eliminate the color shift caused by changes in illumination conditions, further ensuring the color accuracy of the apparent image.

[0028] Step 303: Through this illumination normalization processing, an image that can reflect the essential characteristics of skin lesions can be obtained.

[0029] The specific steps of step S40 include:

[0030] Step 401: Calculate the color representation center of the skin lesion area in the CIE L*a*b* color space based on the skin appearance image with unchanged illumination obtained in step S30. The CIE L*a*b* color space is a perceptually uniform color space, where L represents brightness, and a and b* represent chromaticity, respectively.

[0031] Step 402: By calculating the Lab* value of the lesion area in the color space, a color representation center that can quantitatively describe the degree of the lesion can be obtained.

[0032] The specific steps of step S50 include:

[0033] Step 501: Mapping the color representation center of the skin lesion degree obtained in step S40 in a dark box environment under a standard white light source to obtain a corrected color center.

[0034] Step 502: The purpose of this step is to eliminate the color deviation caused by different lighting conditions, so that the representation of the color center is more accurate and reliable. The standard white light source in a darkroom environment can provide stable and controllable lighting conditions.

[0035] The specific steps of step S60 include:

[0036] Step 601: Obtain the standard value of the color representation center of each skin lesion degree (such as severe, moderate, mild, and cured) in a dark box environment, and record it as the lesion degree color standard value.

[0037] Step 602: Calculate the degree of proximity between the corrected color center obtained in step S50 and the color standard value of each lesion degree, that is, calculate the Euclidean distance between them.

[0038] Step 603: Select the lesion degree corresponding to the lesion degree color standard value close to the maximum probability as the current lesion degree of the skin area to be evaluated.

[0039] The formula for calculating the lesion area ratio is:

[0040] The formula for calculating the lesion color difference is:

[0041]

[0042] in The Lab value for healthy skin. is the Lab value of the lesion area.

[0043] The formula for calculating the lesion severity index is:

[0044]

[0045] Among them, w1 and w2 are weight coefficients, which can be adjusted according to the specific disease type, and w1+w2=1.

[0046] The formula for calculating the rate of change of the lesion severity index is:

[0047]

[0048] The efficacy level is determined according to the change rate and expressed in logical language, specifically: if the change rate ≥ 75%; efficacy = "significantly effective";

[0049] elif 50%≤Change rate<75%:Efficacy = "Effective";

[0050] elif 25%≤Change rate<50%:Efficacy="Improvement";

[0051] elif 0%≤Change rate<25%:Efficacy="slightly improved";

[0052] else:Efficacy = "ineffective".

[0053] Furthermore, the severity of the disease includes four levels: severe, moderate, mild, and cured.

[0054] Furthermore, the illumination normalization process includes using homomorphic filtering technology and a color constancy algorithm.

[0055] Compared with the prior art, the method provided by the present invention for evaluating the efficacy of topical medications used by patients with skin diseases makes full use of standardized skin imaging systems and advanced deep learning image processing technology, and can objectively and quantitatively analyze the characteristic indicators of skin lesions. Through a series of preprocessing, segmentation, normalization and other steps, the accuracy and reliability of image data are ensured. On this basis, combined with the standard color library in a darkroom environment, quantitative indicators reflecting the degree of lesions can be calculated, thereby providing a more objective and accurate basis for evaluating the efficacy of treatment. Specifically, the following technical effects are brought about:

[0056] 1. Objective and quantitative analysis of skin lesion characteristics. By using a standardized skin imaging system and a deep learning image segmentation algorithm, the skin lesion area can be accurately extracted and the color representation center of the lesion area can be calculated in the CIE L*a*b* color space. These quantitative indicators can more objectively reflect the extent of skin lesions and overcome the limitations of traditional subjective evaluation methods.

[0057] 2. Eliminate interference from factors such as lighting. The method of the present invention uses homomorphic filtering and color constancy algorithms in the image preprocessing and color normalization steps to effectively reduce color deviations caused by changes in lighting conditions. This ensures the reliability and repeatability of subsequent analysis results.

[0058] 3. Compare and analyze with the standard color library. The present invention compares and analyzes the color representation center of the lesion area with the pre-established standard color library. By calculating the proximity probability, the degree of the current skin lesion can be judged more accurately, providing an objective basis for efficacy evaluation.

[0059] 4. Comprehensive evaluation of multiple skin characteristic indicators. In addition to color characteristics, the present invention also calculates multiple indicators such as skin lesion area ratio and color difference to construct a comprehensive lesion degree index. This multi-dimensional feature fusion can more comprehensively reflect the overall condition of skin lesions.

[0060] 5. Quantitatively track the trend of efficacy changes. By calculating the change rate of the lesion degree index, the method of the present invention can objectively quantify the efficacy level of the treatment plan and give a corresponding grade judgment. This provides a basis for doctors to adjust the treatment plan and helps to improve the pertinence and effectiveness of the treatment.

[0061] In summary, the method for evaluating the efficacy of topical medications for skin disease patients proposed in the present invention fully utilizes the advantages of image processing and data analysis technology to achieve a quantitative description of the state of skin lesions. Compared with the existing subjective evaluation method, this method has better objectivity, repeatability and consistency, and solves the technical problem that the acquisition of existing skin images is often affected by ambient lighting conditions, which affects the accuracy of the evaluation of the efficacy of topical medications. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A flow chart of the method provided by the present invention;

[0063] Figure 2 It is a comparison diagram before and after pretreatment in the embodiment;

[0064] Figure 3 To accurately reflect the essential characteristics of skin lesions;

[0065] Figure 4 is the mask image of the lesion area;

[0066] Figure 5 This is a diagram for evaluating the treatment effect. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0068] like Figure 1 As shown, it is a flow chart of a method for evaluating the efficacy of topical medication for patients with skin diseases provided by the present invention. The method comprises the following steps:

[0069] S10, using a standardized skin imaging system to obtain a high-resolution digital image of the skin area to be evaluated, and preprocessing the digital image to obtain a first image;

[0070] S20, performing semantic segmentation on the first image using a deep learning-based image segmentation algorithm to extract a skin lesion region of interest;

[0071] S30, performing illumination normalization processing on the skin lesion area of ​​interest to obtain an illumination-invariant skin surface image;

[0072] S40, calculating the color representation center of the skin lesion degree in the CIE L*a*b* color space according to the skin appearance image with unchanged illumination;

[0073] S50, mapping the color representation center of the skin lesion degree in a dark box environment under a standard white light source to obtain a corrected color center;

[0074] S60, obtaining the standard value of the color representation center of each skin lesion degree in the darkroom environment, recorded as the lesion degree color standard value; calculating the probability of the corrected color center approaching the color standard value of each lesion degree, and taking the lesion degree corresponding to the lesion degree color standard value with the highest probability of approaching as the current lesion degree of the skin area to be evaluated;

[0075] S70, calculating characteristic indicators of the skin lesion area, including the area ratio, lesion color difference, and lesion severity index;

[0076] S80. Calculate the change rate of the lesion severity index based on the characteristic indicators of the skin lesion area, and determine the efficacy grade based on the change rate.

[0077] The specific implementation methods of the above steps are described in detail below:

[0078] The specific implementation method of step S10 is: first, a high-resolution digital image of the skin area to be evaluated is obtained using a standardized skin imaging system. This imaging system usually includes a professional digital camera, a light source and a fixed shooting platform to ensure that each image collected has consistent lighting conditions and imaging parameters. After obtaining the original digital image, it needs to be preprocessed. Preprocessing includes steps such as color correction, denoising and geometric correction. Color correction aims to eliminate color distortion caused by factors such as uneven lighting or camera color deviation, thereby ensuring the accuracy of image color information. Denoising is aimed at the noise components in the image and uses a common low-pass filtering algorithm to eliminate them. The purpose of geometric correction is to correct the geometric distortion of the image caused by factors such as camera tilt or shooting angle. After these preprocessing steps, a high-quality, interference-free digital image of the skin area can be obtained, laying the foundation for subsequent computational analysis.

[0079] The specific implementation method of step S20 is: the pre-processed first image is subjected to semantic segmentation using an image segmentation algorithm based on deep learning. It is recommended to use common segmentation network models such as U-Net or Mask R-CNN. Through these deep learning-based segmentation algorithms, the skin lesion area of ​​interest can be accurately extracted from the overall image. It is worth noting that these deep learning segmentation algorithms need to be trained in advance using a large number of labeled skin images to improve the accuracy of identifying the lesion area. After this step, an accurate mask image of the skin lesion area can be obtained.

[0080] The specific implementation of step S30 is: perform illumination normalization processing on the extracted skin lesion area. Homomorphic filtering technology and color constancy algorithm can be used here. Homomorphic filtering can effectively suppress the non-uniform illumination components in the image, thereby obtaining an apparent skin image with constant illumination. The color constancy algorithm can eliminate the color shift caused by changes in illumination conditions, and further ensure the color accuracy of the apparent image. Through this illumination normalization processing, an image that can reflect the essential characteristics of skin lesions can be obtained.

[0081] The specific implementation of step S40 is: according to the skin appearance image with unchanged illumination obtained in step S30, the color representation center of the skin lesion area is calculated in the CIEL*a*b* color space. The CIE L*a*b* color space is a perceptually uniform color space, where L represents brightness, and a and b represent chromaticity respectively. By calculating the Lab value of the lesion area in the color space, a color representation center that can quantitatively describe the degree of the lesion can be obtained.

[0082] The specific implementation of step S50 is: the color representation center of the skin lesion degree obtained in step S40 is mapped in a dark box environment under a standard white light source to obtain a corrected color center. The purpose of this step is to eliminate the color deviation caused by different lighting conditions, so that the representation of the color center is more accurate and reliable. The standard white light source in a dark box environment can provide stable and controllable lighting conditions, which is conducive to obtaining a corrected color center value.

[0083] The specific implementation of step S60 is: first, obtain the standard value of the color representation center of each skin lesion degree (such as severe, moderate, mild, and cured) in the dark box environment, and record it as the lesion degree color standard value. Then, calculate the degree of proximity between the corrected color center obtained in step S50 and the color standard value of each lesion degree, that is, calculate the Euclidean distance between them. Next, select the lesion degree corresponding to the lesion degree color standard value with the highest probability of proximity as the current lesion degree of the skin area to be evaluated. The core idea of ​​this step is to infer the lesion degree of the current skin condition by comparing the actual color of the skin area with the color of each lesion degree in the standard color library.

[0084] The specific implementation of step S70 is: calculating three characteristic indicators of the skin lesion area, namely, the variable area ratio, the lesion color difference and the lesion degree index. The calculation formula of the variable area ratio is: This indicator reflects the proportion of the lesion area relative to the entire skin area. The calculation formula for the lesion color difference is: in The Lab value for healthy skin. is the Lab value of the lesion area. This index describes the difference between the lesion area and healthy skin in the color space. The calculation formula of the lesion degree index is: Where w1 and w2 are adjustable weight coefficients, and w1+w2 = 1. This index takes into account two factors: the proportion of lesion area and the difference in lesion color, and can more comprehensively reflect the overall degree of skin lesions.

[0085] The specific implementation of step S80 is: according to the three skin lesion area characteristic indicators calculated in step S70, further calculate the change rate of the lesion degree index. The calculation formula is: By analyzing the change rate, the efficacy level of the current treatment can be determined. The specific rules are as follows: if the change rate is greater than or equal to 75%, it is judged as "significantly effective"; if the change rate is between 50% and 75%, it is judged as "effective"; if the change rate is between 25% and 50%, it is judged as "improved"; if the change rate is between 0% and 25%, it is judged as "slightly improved"; if the change rate is less than 0%, it is judged as "ineffective". The purpose of this step is to comprehensively analyze multiple quantitative indicators of skin lesions, so as to objectively evaluate the efficacy of the current treatment plan.

[0086] In general, the method for evaluating the efficacy of smeared medication for patients with skin diseases proposed in the present invention makes full use of the standardized skin imaging system and the image processing technology based on deep learning, and can quantitatively analyze the characteristics of the area, color difference and overall degree of skin lesions. Through a series of preprocessing steps such as correction, segmentation, and normalization, the accuracy and reliability of the image data are ensured. On this basis, combined with the standard color library under the dark box environment, the characteristic index that objectively reflects the degree of the lesion is calculated, and the efficacy of the treatment plan is analyzed accordingly. This quantitative evaluation method based on image analysis is more accurate, objective and repeatable than traditional subjective judgment. It helps doctors better monitor and guide the treatment process of patients with skin diseases and improve clinical efficacy.

[0087] Specifically, the principle of the present invention is to combine a standardized skin imaging system with advanced deep learning image processing technology to objectively and quantitatively analyze and evaluate skin lesions. Specifically, the principle of its technical solution is as follows:

[0088] First, a standardized skin imaging system is used to obtain high-resolution digital images of the skin area to be evaluated. This imaging system can provide stable and controllable lighting conditions to ensure that each image acquired has consistent imaging parameters. Then, the acquired original image is preprocessed by color correction, denoising, and geometric correction to eliminate interference caused by factors such as uneven lighting and camera color deviation, and obtain a high-quality skin area image.

[0089] Next, use a deep learning-based image segmentation algorithm, such as U-Net or Mask R-CNN, to accurately extract the skin lesion area of ​​interest from the overall image. This deep learning segmentation algorithm needs to be trained with a large number of labeled skin images in advance to improve the accuracy of identifying the lesion area. The segmentation algorithm can obtain an accurate mask image of the skin lesion area.

[0090] In order to eliminate the influence of illumination factors on color information, the present invention uses homomorphic filtering and color constancy algorithm to perform illumination normalization processing on the extracted lesion area. Homomorphic filtering can effectively eliminate the non-uniform illumination components in the image, and the color constancy algorithm can eliminate the color shift caused by changes in illumination conditions. After these processes, an image that can accurately reflect the essential characteristics of skin lesions can be obtained.

[0091] On this basis, the present invention uses the CIE L*a*b* color space to calculate the color representation center of the skin lesion area. CIE L*a*b* is a perceptually uniform color space, where L represents brightness, and a and b represent chromaticity. By calculating the Lab value of the lesion area in this color space, a color representation center that can quantitatively describe the degree of the lesion can be obtained.

[0092] In order to further eliminate the color deviation under different lighting conditions, the present invention maps the above color representation center in a dark box environment under a standard white light source to obtain a corrected color center. This step ensures that the representation of the color center is more accurate and reliable.

[0093] Next, the present invention obtains the standard value of the color representation center of each skin lesion degree (such as severe, moderate, mild, and cured) under the dark box environment, and calculates the Euclidean distance between the corrected color center of the current skin area and each standard value. By analyzing the lesion degree corresponding to the standard value with the highest probability of approaching, the current lesion degree of the skin area to be evaluated can be determined.

[0094] In addition to color features, the present invention also calculates other characteristic indicators such as the area ratio of skin lesions and lesion color differences, and constructs a comprehensive lesion degree index. This index can more comprehensively reflect the overall condition of skin lesions.

[0095] Finally, the present invention provides a basis for doctors to evaluate the effect of treatment plans by calculating the change rate of the lesion degree index and determining the efficacy level according to the change rate. By tracking the change trend of the lesion degree index, the treatment effect can be judged more timely and accurately, providing support for adjusting the treatment plan.

[0096] In general, the technical solution of the present invention makes full use of image processing and data analysis technology to achieve an objective and quantitative description of the state of skin lesions. Compared with the traditional subjective evaluation method, this method has better repeatability and consistency, provides doctors with a more scientific and reliable basis, and helps to improve the level of clinical diagnosis and treatment.

[0097] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: A 40-year-old psoriasis patient, Mr. Zhang, was admitted to a dermatology clinic. At the initial visit, Mr. Zhang reported that plaque-like erythema appeared on the inner side of his legs, and the skin lesions were significantly aggravated compared with the previous week, accompanied by mild itching. After preliminary visual examination, the doctor determined that this was a typical acute attack of psoriasis. In order to objectively evaluate the state of Mr. Zhang's skin lesions, the doctor decided to use the skin disease efficacy evaluation method proposed in the present invention.

[0098] First, the doctor arranged the patient to sit in front of a standardized skin imaging system to collect high-resolution digital images of the skin area to be evaluated (the inner side of the leg). The imaging system consists of a professional digital camera, a spotlight, and a fixed shooting platform to ensure that each image collected has consistent imaging parameters. By preprocessing the original image with color correction, denoising, and geometric correction, a high-quality skin area image Igeo_cor is finally obtained. Figure 2 As shown, Figure 2 The left side shows the image of the lesion location taken by Mr. Zhang at home, and the right side shows the preprocessed image. In order to eliminate the influence of light factors on color information, the doctor then normalizes the light of the extracted lesion area. Specifically, the homomorphic filtering and color constancy algorithm are used to obtain an image Icolor_const that can accurately reflect the essential characteristics of skin lesions. Figure 3 Next, the doctor used the U-Net segmentation algorithm based on deep learning to perform semantic segmentation on Igeo_cor, accurately extract the skin lesion area, and set it to black and white contrast as a mask. The resulting mask image is as follows: Figure 4 Using the mask and Icolor_const superposition, the doctor uses the CIE L*a*b* color space to calculate the color representation center of the lesion area These color feature values ​​can quantitatively reflect the extent of skin lesions.

[0099] In order to further eliminate the color deviation under different lighting conditions, the doctor mapped the above color center in a dark box environment under a standard white light source to obtain the corrected color center. After this step, the representation of color center is more accurate and reliable.

[0100] Next, the doctor obtained the standard values ​​of the color representation center of four typical skin lesion degrees (severe, moderate, mild, and cured) under a dark box environment, as shown in Table 1.

[0101] Table 1 Standard color center values ​​for different lesion degrees

[0102]

[0103] The doctor then calculated the corrected color center of Mr. Zhang's current skin area. And compared with the above standard value, it is found that it is closest to the standard value of moderate lesion degree. Therefore, it is preliminarily judged that Mr. Zhang's current skin lesion degree is moderate.

[0104] In addition to the color feature, the doctor also calculated two other feature indicators:

[0105] 1. Lesion area ratio A r : According to the segmentation results, the area of ​​the lesion on the inner side of Mr. Zhang's leg is A lesion =12.4cm 2 , total assessment area A total =82.6cm 2 ,therefore

[0106] 2. Lesion color difference ΔE: Calculate the color of the lesion area Healthy skin color The difference,

[0107] Based on the above three characteristic indicators, the doctor calculated Mr. Zhang's current lesion severity index as:

[0108]

[0109] Among them, w1=0.6, w2=0.4 are the preset weight coefficients.

[0110] This disease severity index I s =0.27 indicates that Mr. Zhang’s current skin lesion condition is at a moderate level.

[0111] In order to evaluate the treatment effect of Mr. Zhang, the doctor recorded the lesion severity index at the last evaluation. According to the method of the present invention, the change rate of the current lesion severity index is calculated as:

[0112]

[0113] Based on the rate of change r=22.9%, the doctor determined that Mr. Zhang’s current treatment effect was “improved”.

[0114] In order to further analyze the treatment effect, the doctor drew a trend chart of the change in the lesion severity index, such as Figure 5 As shown in the figure, after continuous treatment, Mr. Zhang's skin lesions showed a trend of gradual improvement. This provides a basis for doctors to adjust the treatment plan.

[0115] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating the efficacy of topical medications used by patients with skin diseases, characterized in that: The following steps are involved: S10, using a standardized skin imaging system to obtain a high-resolution digital image of the skin area to be evaluated, and preprocessing the digital image to obtain a first image; S20, performing semantic segmentation on the first image using a deep learning-based image segmentation algorithm to extract a skin lesion area of ​​interest; S30, performing illumination normalization processing on the skin lesion region of interest to obtain an illumination-invariant skin surface image; S40, calculating the color representation center of the skin lesion degree in the CIE L*a*b* color space according to the skin appearance image with unchanged illumination; S50, mapping the color representation center of the skin lesion degree in a dark box environment under a standard white light source to obtain a corrected color center; S60, obtaining the standard value of the color representation center of each skin lesion degree in the dark box environment, recorded as the lesion degree color standard value; calculating the proximity probability between the corrected color center and each lesion degree color standard value, and taking the lesion degree corresponding to the lesion degree color standard value with the highest proximity probability as the current lesion degree of the skin area to be evaluated; S70, calculating characteristic indicators of the skin lesion area, including the area ratio, lesion color difference, and lesion severity index; S80. Calculate the change rate of the lesion severity index according to the characteristic index of the skin lesion area, and determine the efficacy grade according to the change rate.

2. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 1, characterized in that: The preprocessing includes color correction, denoising and geometric correction.

3. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 2, characterized in that: The image segmentation algorithm based on deep learning adopts U-Net or Mask R-CNN.

4. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 1, characterized in that: The formula for calculating the lesion area ratio is:

5. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 4, characterized in that: The formula for calculating the lesion color difference is: in The Lab value for healthy skin. is the Lab value of the lesion area.

6. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 5, characterized in that: The formula for calculating the lesion severity index is: Where w1 and w2 are weight coefficients.

7. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 6, characterized in that: The formula for calculating the rate of change of the lesion severity index is:

8. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 7, characterized in that: The efficacy level is determined according to the change rate. Specifically, if the change rate is greater than or equal to 75%, it is judged as "remarkably effective"; if the change rate is between 50% and 75%, it is judged as "effective"; if the change rate is between 25% and 50%, it is judged as "improved"; If the rate of change is between 0% and 25%, it is considered "slight improvement"; If the change rate is less than 0%, it is judged as "invalid".

9. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 8, characterized in that: The severity of the disease includes four levels: severe, moderate, mild, and cured.

10. A method for evaluating the efficacy of topical medication for patients with skin diseases according to claim 9, characterized in that: The illumination normalization process includes using a homomorphic filtering technique and a color constancy algorithm.

Images