System and method for evaluating severity of psoriasis
The integration of Trans-Unet and ConvNeXt models with improved loss functions addresses inefficiencies in psoriasis scoring, providing an objective and accurate psoriasis severity assessment system.
Patent Information
- Application Number
- CN202411119835.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Current psoriasis severity assessment methods rely heavily on visual observation, leading to high workload, low accuracy, non-objective scoring, and lack of standardization, resulting in inefficient and inconsistent evaluations.
A psoriasis severity assessment system integrating an image segmentation module, psoriasis skin lesion surface area measurement module, and psoriasis severity assessment module, utilizing the Trans-Unet and ConvNeXt models for intelligent image segmentation and severity assessment, with improved cross-entropy and contrastive loss functions for training.
The system reduces medical staff workload, enhances evaluation efficiency, improves accuracy, and ensures objective, repeatable, and interpretable psoriasis severity assessments.
Smart Images

Figure CN119069087A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and more particularly to a system and method for evaluating the severity of psoriasis. Background Art
[0002] Psoriasis is a chronic, recurrent and inflammatory skin disease induced by multiple factors. Its typical clinical manifestations are scaly erythema or plaques. It can cause multiple complications such as metabolic syndrome, cardiovascular disease and inflammatory arthritis, seriously affecting the quality of life of patients.
[0003] At present, the PASI score is an important standard for assessing the severity of psoriasis, which mainly includes the lesion area score and the lesion severity score. The lesion area score is mainly based on the area of the affected skin, and the lesion severity score is mainly based on the degree of erythema, the degree of infiltration, and the degree of desquamation. At present, when dermatologists perform PASI scoring, they mainly rely on visual observation for the above four indicators, which mainly has problems such as large workload, low work efficiency and high cost. There are also problems such as the scoring results are not objective enough, the evaluation accuracy is low, the interpretability and replicability are poor, and the scoring standards are not unified. Summary of the invention
[0004] In view of this, the present invention uses computer deep learning technology to provide a psoriasis severity assessment system and method that at least solves some of the above-mentioned technical problems. In the present invention, an image segmentation module, a psoriasis skin lesion surface area measurement module and a psoriasis severity assessment module are integrated together, and the Trans-Unet model and the ConvNeXt model are used to facilitate the intelligent completion of image segmentation and psoriasis severity assessment tasks, thereby reducing the workload of medical staff, improving work efficiency, reducing costs, and improving assessment accuracy, and has the advantages of objectivity, explainability and repeatability.
[0005] To achieve the above object, the technical solution adopted by the present invention is:
[0006] In a first aspect, the present invention provides a psoriasis severity assessment system, the system comprising: an image segmentation module, a psoriasis skin lesion surface area measurement module and a psoriasis severity assessment module, wherein:
[0007] The image segmentation module is used to extract global information of the input image based on the Trans-Unet model and perform pixel-level segmentation on the input image;
[0008] The psoriasis skin lesion body surface area measurement module is used to calculate the skin lesion area ratio according to the image segmentation result, and obtain the psoriasis severity level of each image by embedding the set threshold; and generate pseudo-label information for auxiliary learning of the psoriasis severity assessment module;
[0009] The psoriasis severity assessment module is used to train training samples and pseudo-label information based on the convolutional neural network ConvNeXt, construct a psoriasis severity assessment model through comparative learning, and train and optimize the psoriasis severity assessment model through improved cross entropy loss function and contrast loss function; and use the optimized psoriasis severity assessment model to perform psoriasis severity assessment.
[0010] Furthermore, in the Trans-Unet model, the Transformer network is used to extract the global information of the image, and the U-net network is used to up-sample and down-sample the target area to achieve pixel-level segmentation of the image.
[0011] Furthermore, the psoriasis lesion surface area measurement module calculates the diseased area P of each segmented image according to the image segmentation result. i and the area of normal skin area Q i , through the diseased area P i and the area of normal skin area Q i The calculation formula for the skin lesion area ratio is:
[0012]
[0013] In the formula, B i Indicates the percentage of skin lesion area.
[0014] Furthermore, in the convolutional neural network ConvNeXt, depth-wise separable convolution is used to split a full convolution into convolution within the channel and convolution between channels.
[0015] Furthermore, in the convolutional neural network ConvNeXt, feature extraction is divided into four stages, each stage consists of repeatedly stacked dense computing units, and the ratio of dense computing units in the four stages is 3:3:9:3.
[0016] Furthermore, in the convolutional neural network ConvNeXt, LN is used to normalize the features.
[0017] Furthermore, in the convolutional neural network ConvNeXt, the GELU activation function is used to increase the nonlinear fitting ability of the network.
[0018] Furthermore, the improved cross entropy loss function is calculated as follows:
[0019]
[0020] p ic =p(y ic |x i)
[0021] p' ic =p(y′ ic |x i )
[0022] In the formula, N represents the number of batch samples, t represents the number of categories, and x i Represents the input sample, y ic Indicates the label, y' ic represents the pseudo label, p ic Represents x i The label is y ic The probability of ic Represents x i The label is y' ic The probability of time.
[0023] Furthermore, the contrast loss function is calculated as follows:
[0024]
[0025] In the formula, z i ,z j is a pair of positive samples, z i ,z K are each other’s negative samples, and τ is a temperature parameter.
[0026] In a second aspect, the present invention further provides a method for evaluating the severity of psoriasis, which is applied to the above-mentioned system for evaluating the severity of psoriasis to evaluate the severity of psoriasis, and the method comprises:
[0027] Input training image data and pseudo-label information of psoriasis severity;
[0028] Perform data enhancement on the training data to generate two sets of data;
[0029] The two sets of data are sent to the convolutional neural network ConvNeXt respectively to form a contrastive learning framework, defining the data of the same category as positive samples and the data of other categories as negative samples;
[0030] The model parameters are set and trained through the joint optimization objectives to construct a psoriasis severity assessment model; and the psoriasis severity assessment model is used to conduct psoriasis severity assessment.
[0031] Compared with the prior art, the present invention has at least the following beneficial technical effects:
[0032] The present invention provides a psoriasis severity assessment system and method. The present invention integrates an image segmentation module, a psoriasis skin lesion surface area measurement module, and a psoriasis severity assessment module, and uses a Trans-Unet model and a ConvNeXt model to intelligently complete image segmentation and psoriasis severity assessment tasks, thereby reducing the workload of medical staff, improving work efficiency, and reducing costs. In addition, the present invention introduces a comparative learning method to improve the classification accuracy, robustness, and generalization ability of the model, which helps to improve the accuracy of psoriasis severity assessment and has the advantages of objectivity, interpretability, and repeatability.
[0033] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.
[0034] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0036] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.
[0037] Figure 1 A schematic diagram of the structure of a psoriasis severity assessment system provided by an embodiment of the present invention.
[0038] Figure 2 A schematic diagram of the network structure of the Trans-Unet model provided in an embodiment of the present invention.
[0039] Figure 3 A schematic diagram of the network structure of the ConvNeXt model provided in an embodiment of the present invention.
[0040] Figure 4 A schematic flow chart of a method for assessing the severity of psoriasis provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0042] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0043] See also Figure 1 As shown, an embodiment of the present invention provides a psoriasis severity assessment system, which includes: an image segmentation module, a psoriasis skin lesion surface area measurement module and a psoriasis severity assessment module. The present invention combines the three modules in a single framework to achieve accurate, fast and intelligent psoriasis severity assessment. The technologies involved include Trans-Unet model and ConvNeXt model, etc., which are used to complete image segmentation and psoriasis severity assessment tasks.
[0044] The specific implementation and related working principle of a psoriasis severity assessment system of the present invention are described in detail below:
[0045] (1) Image Segmentation Module
[0046] U-net is one of the earlier image segmentation algorithms that used fully convolutional networks. It is an improved semantic segmentation network based on FCN. Compared with FCN networks, U-net has a simple network structure and significant experimental results. It can retain more local details in the segmentation results and is widely used in semantic segmentation scenarios such as medical imaging and drone remote sensing imaging. The U-net network achieves pixel-level segmentation of images by upsampling and downsampling the target area. However, due to the U-net network's translation invariance and lack of ability to capture long-term dependencies, it requires multiple layers of convolution to obtain abstract global information, which affects the efficiency of image segmentation. As a model for natural language processing tasks, the Transformer network can use the long-term dependencies between pixels to effectively extract global information from images, but it has poor results in extracting detailed information.
[0047] In this embodiment, the present invention preferably uses the Trans-Unet network model for image segmentation. The Trans-Unet network model structure is as follows: Figure 2As shown in Figure 1, part (a) is a schematic diagram of a layer of Transformer structure, embedding sequence After normalization, multi-head attention and multi-layer perceptron, the output information Z1. Part (b) is the complete Trans-Unet network architecture. The encoder part first performs three-layer convolution downsampling on the input image, embeds the image block into the feature map obtained by CNN, and then inputs the one-dimensional vector after block embedding into the 12-layer Transfomer structure. In the present invention, Trans-Unet introduces Transformer into U-net, and has the advantages of Transformer and U-net. The Transformer is added to the encoder part to model the global information of the image, which greatly improves the segmentation accuracy.
[0048] (2) Psoriasis lesion body surface area measurement module
[0049] The present invention calculates the diseased area P of each segmented image according to the image segmentation results. i and the area of normal skin area Q i According to the formula Calculate the proportion of the diseased area B i .
[0050] In this embodiment, psoriasis can be preferably divided into mild, moderate and severe according to the proportion of skin lesions of the patient. Therefore, by embedding the set threshold in this module, the severity level of psoriasis for each picture can be obtained. Secondly, using this information, the module generates it as pseudo-label information for auxiliary learning of the psoriasis severity assessment module. Pseudo-labels can not only correct the inaccuracy of manual labeling, but also provide a valuable, referenceable and interpretable information for unlabeled data, making up for the inexplicable shortcomings of deep learning.
[0051] (3) Psoriasis severity assessment module
[0052] In this embodiment, the overall structure of the convolutional neural network ConvNeXt model used in the present invention is as follows: Figure 3 As shown. After a batch of images is input, it first passes through the Stem network, adjusts the size and performs dimensionality reduction, and then enters the 4 stages of the backbone network. Each stage is composed of repeated block units, but the first block in each stage is responsible for the downsampling task, so the step size is set to 2 at this time. The original convolutional neural network uses full-channel convolution to store a large number of weight parameters, thereby realizing the intensive calculation process of recognizing and selecting image features, but this process is very resource-intensive, so the present invention preferably uses deep separable convolution in the network, splitting a full convolution into convolution within the channel and convolution between channels, which greatly reduces the amount of calculation.
[0053] In this embodiment, in the four stages of feature extraction, each stage consists of repeatedly stacked dense computing units, and the design structure of these units is basically the same. CNN generally adopts a ratio of 3:4:6:3, but ConvNeXt adjusts this ratio to 3:3:9:3 to achieve higher performance. This modification increases the overall parameter volume, improves the calculation speed, and gives the network a more powerful information representation capability.
[0054] In this embodiment, the present invention draws on a lot of ideas of Transformer in the selection of some hyperparameters and the optimization of the network forward reasoning process. Although each improvement is very small, the overall result is good. Grouped convolution is used to reduce the amount of network calculation. At the same time, layer normalization (LN) is used instead of batch normalization (BN). Because normalization has no learnable parameters in the network, it generally exists as a bottom line in the network operation process, which can effectively prevent the increase in the number of network layers from causing the offset of the feature map of the middle layer and the negative effects such as gradient disappearance and gradient explosion. BN and LN are also normalization of feature maps, both of which calculate the mean and variance of all features, but the dimensions are different. BN is normalized for the feature maps between channels on the same batch, and LN is normalized for the other three dimensional information between batch dimensions. The GELU activation function is used in the present invention, and its main function is to increase the nonlinear fitting ability of the network and enable the network to have a larger information space.
[0055] In this embodiment, the psoriasis severity assessment module: uses a convolutional neural network ConvNeXt to train training samples and severity assessment labels (pseudo labels) to build a psoriasis severity assessment model.
[0056] The core idea of the contrastive learning (CL) algorithm is to use the sample's own information to generate supervision signals, and to extract better modal features by continuously narrowing the distance between samples of the same type and pushing the distance between different types farther. In existing contrastive learning research, there are two main ways to measure the distance between features: one is to optimize the network by minimizing the Euclidean distance between samples of the same type, and the other is to optimize the network by maximizing the cosine similarity between the features of samples of the same type and the features of samples of different types. Among them, the Euclidean distance is the most intuitive measurement method.
[0057] During the training process, the loss function is a function used to measure the difference between the model's predicted output and the true value. The model is trained and optimized by minimizing the loss function. Contrastive loss and cross entropy loss are two commonly used loss functions in deep learning. The cross entropy loss function is mainly used for classification problems to measure the accuracy of the model in classifying samples; the contrastive loss function is mainly used to measure the similarity and difference between samples. The combination of the two loss functions can more effectively improve the classification accuracy, robustness, and generalization ability of the model.
[0058] In a specific embodiment, the calculation formula of the improved cross-entropy (CE) loss function is as follows:
[0059]
[0060] p ic =p(y ic |x i )
[0061] p′ ic =p(y′ ic |x i ) (2)
[0062] Among them, N is the number of batch samples, t is the number of categories, and x i is the input sample (i.e. each segmented image), y ic is the label, y' ic is a pseudo label, p ic Represents x i The label is y i The probability of ic Represents x i The label is y' ic The essence of the cross entropy loss function comes from the concept of information entropy in information theory, which can be understood as the difference function between the predicted probability distribution and the true probability distribution.
[0063] The calculation formula of contrast loss is shown in (3):
[0064]
[0065] The cosine similarity is used to calculate the distance between two samples u and v. The formula is shown in (4):
[0066]
[0067] Among them, z i ,z j is a pair of positive samples, z i ,z Kare negative samples of each other, and τ is a temperature parameter. The above loss function (3) is calculated for each pair of images in the batch training of the deep learning model, and finally the sum of all loss functions is averaged, which is the final contrast loss function value, as shown in formula (5):
[0068]
[0069] During the training phase, the model can use labels for supervised training to ensure its discrimination ability, and on the other hand, it can use contrastive learning loss function to learn more generalized representations. Therefore, the final overall optimization strategy is shown in Formula 6:
[0070] L=L CE +α*L CL (6)
[0071] Here, α represents a hyperparameter.
[0072] From the description of the above embodiments, those skilled in the art can know that the present invention provides a psoriasis severity assessment system, which integrates an image segmentation module, a psoriasis skin lesion surface area measurement module, and a psoriasis severity assessment module, and uses a Trans-Unet model and a ConvNeXt model to intelligently complete image segmentation and psoriasis severity assessment tasks, thereby improving work efficiency and reducing costs. The present invention can achieve rapid, objective, and consistent evaluation, which helps to improve computer-aided systems and reduce doctors' workload.
[0073] Further, refer to Figure 4 As shown, an embodiment of the present invention further provides a method for evaluating the severity of psoriasis, which is applied to a psoriasis severity evaluation system of the above embodiment to evaluate the severity of psoriasis. The method comprises:
[0074] a) Input training image data and pseudo label information of psoriasis severity.
[0075] b) Perform data enhancement on the training data to generate two sets of data. Preferably, the training set data processing includes normalization processing, and constructing positive and negative samples through visual semantic transformation, such as jigsaw puzzles, cropping, blurring, color transformation, etc. to construct new samples, and these samples are still considered to be of the same category. Different from the existing contrastive learning, the present invention proposes a novel positive and negative sample construction method.
[0076] c) The two sets of data are respectively sent to a double-branch network with upper and lower symmetry (i.e., convolutional neural network ConvNeXt) to form a contrastive learning framework, defining data of the same category as positive samples and data of other categories as negative samples.
[0077] d) Setting model parameters and training through joint optimization objectives to construct a psoriasis severity assessment model and perform psoriasis severity assessment.
[0078] The method for assessing the severity of psoriasis provided in the embodiment of the present invention introduces a comparative learning method to improve the classification accuracy, robustness and generalization ability of the model, which helps to improve the accuracy of the assessment of the severity of psoriasis and has the advantages of objectivity, interpretability and repeatability. The implementation principle and the technical effects produced refer to the aforementioned system embodiment. For the sake of brief description, the parts not mentioned in this embodiment can also refer to the corresponding contents in the aforementioned system embodiment, which will not be repeated here.
[0079] In addition, an embodiment of the present invention further provides a storage medium on which one or more programs readable by a computing device are stored. The one or more programs include instructions. When the instructions are executed by the computing device, the computing device executes a psoriasis severity assessment method in the above embodiment.
[0080] Examples of computer-readable storage media include: read-only memory (ROM), random-access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random-access memory (RAM), dynamic random-access memory (DRAM), static random-access memory (SRAM), flash memory, non-volatile memory, CD-ROM, DVD-ROM, Blu-ray or optical disk storage, hard disk drive (HDD), solid-state drive (SSD), card-type memory (such as a multimedia card, a secure digital (SD) card, or an extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state disk, and any other device configured to store a computer program and any associated data, data files, and data structures in a non-transitory manner and provide the computer program and any associated data, data files, and data structures to a processor or computer so that the processor or computer can execute the computer program. The computer program in the above-mentioned computer-readable storage medium can be executed in an environment deployed in a computer device such as a client, a host, an agent device, a server, etc. In addition, in one example, the computer program and any associated data, data files, and data structures are distributed on a networked computer system, so that the computer program and any associated data, data files, and data structures are stored, accessed, and executed in a distributed manner by one or more processors or computers.
[0081] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as systems, methods or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0082] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several distinct components, and by means of a suitably programmed computer.
[0083] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0084] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A psoriasis severity assessment system, characterized in that: The system includes: an image segmentation module, a psoriasis skin lesion surface area measurement module and a psoriasis severity assessment module, wherein: The image segmentation module is used to extract global information of the input image based on the Trans-Unet model and perform pixel-level segmentation on the input image; The psoriasis skin lesion body surface area measurement module is used to calculate the skin lesion area ratio according to the image segmentation result, and obtain the psoriasis severity level of each image by embedding the set threshold; and generate pseudo-label information for auxiliary learning of the psoriasis severity assessment module; The psoriasis severity assessment module is used to train training samples and pseudo-label information based on the convolutional neural network ConvNeXt, construct a psoriasis severity assessment model through comparative learning, and train and optimize the psoriasis severity assessment model through improved cross entropy loss function and contrast loss function; and use the optimized psoriasis severity assessment model to perform psoriasis severity assessment.
2. A psoriasis severity assessment system according to claim 1, characterized in that: In the Trans-Unet model, the Transformer network is used to extract the global information of the image, and the U-net network is used to up-sample and down-sample the target area to achieve pixel-level segmentation of the image.
3. A psoriasis severity assessment system according to claim 1, characterized in that: The psoriasis skin lesion surface area measurement module calculates the diseased area P of each segmented image according to the image segmentation results. i and the area of normal skin area Q i , through the diseased area P i and the area of normal skin area Q i The calculation formula for the skin lesion area ratio is: In the formula, B i Indicates the percentage of skin lesion area.
4. A psoriasis severity assessment system according to claim 1, characterized in that: In the convolutional neural network ConvNeXt, depth-wise separable convolution is used to split a full convolution into convolution within the channel and convolution between channels.
5. A psoriasis severity assessment system according to claim 1, characterized in that: In the convolutional neural network ConvNeXt, feature extraction is divided into four stages. Each stage consists of repeatedly stacked dense computing units. The ratio of dense computing units in the four stages is 3:3:9:
3.
6. A psoriasis severity assessment system according to claim 1, characterized in that: In the convolutional neural network ConvNeXt, LN is used to normalize the features.
7. A psoriasis severity assessment system according to claim 1, characterized in that: In the convolutional neural network ConvNeXt, the GELU activation function is used to increase the nonlinear fitting ability of the network.
8. A psoriasis severity assessment system according to claim 1, characterized in that: The improved cross entropy loss function is calculated as follows: p ic =p(y ic |x i ) p’ ic =p(y′ ic |x i ) In the formula, N represents the number of batch samples, t represents the number of categories, and x i Represents the input sample, y ic Indicates the label, y' ic represents the pseudo label, p ic Represents x i The label is y ic The probability of ic Represents x i The label is y' ic The probability of time.
9. A psoriasis severity assessment system according to claim 1, characterized in that: The contrast loss function is calculated as follows: In the formula, z i ,z j is a pair of positive samples, z i ,z K are each other’s negative samples, and τ is a temperature parameter.
10. A method for assessing the severity of psoriasis, characterized in that: A psoriasis severity assessment system according to any one of claims 1 to 9 is used to assess the severity of psoriasis, the method comprising: Input training image data and pseudo-label information of psoriasis severity; Perform data enhancement on the training data to generate two sets of data; The two sets of data are sent to the convolutional neural network ConvNeXt respectively to form a contrastive learning framework, defining the data of the same category as positive samples and the data of other categories as negative samples; The model parameters are set and trained through the joint optimization objectives to construct a psoriasis severity assessment model; and the psoriasis severity assessment model is used to conduct psoriasis severity assessment.
Citation Information
Patent Citations
Intelligent decision-making system for evaluating severity of psoriasis based on skin image
CN114299068A
Image-based psoriasis severity assessment system
CN114882018A
Image segmentation method and system based on segmentation model, electronic equipment and medium
CN115965631A
Psoriasis assessment system based on skin image and skin electric signal
CN117958755A
Large receptive field attention-enhanced breast ultrasound image segmentation method
CN118261925A
Cited By
Machine vision-based psoriasis example feature recognition method and device
CN119863439A
A method and device for identifying psoriasis case features based on machine vision
CN119863439B
Intelligent staging diagnosis and evaluation method and system for psoriasis
CN122455305A
