Image processing method and device, electronic equipment and storage medium

Abstract

The invention discloses an image processing method and device. The method comprises the following steps: acquiring images of N1 parts of a target object; for each part in the N1 parts, carrying out image segmentation on the image of the part to obtain image data of a focus area of the part; the focus area is a characteristic dermatitis area, and each of the N1 parts comprises at least one focus area; for each part in the N1 parts, training image data of a focus area of the part by using a focus classification model to obtain attribute information and score information corresponding to each focus area of the part; the attribute information comprises at least one of erythema, papule, exfoliation of epidermis and moss change; the score information is one of preset N2 score values; the severity of the skin lesion of the focus area corresponding to each score value in the N2 score values is different. The method and the device can provide reference for a doctor to classify and identify the AD of each part of the body of the target object.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular to an image processing method, device, electronic device and storage medium. Background Art

[0002] Atopic dermatitis (AD) is a common chronic inflammatory, recurrent skin disease that affects the patient's physical and mental health and reduces their quality of life. Currently, the classification and identification of AD mainly rely on the doctor's clinical experience, which is subjective, and the evaluation process is complicated and takes a long time. Summary of the invention

[0003] To solve the above problems, the embodiments of the present application provide an image processing method, device, electronic device and storage medium.

[0004] In a first aspect, an embodiment of the present application provides an image processing method, the method comprising:

[0005] Acquire images of N1 parts of the target object; wherein N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image;

[0006] For each of the N1 parts, image segmentation is performed on the image of the part to obtain image data of the lesion area of ​​the part; the lesion area is an idiopathic dermatitis area, and each of the N1 parts includes at least one lesion area;

[0007] For each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part;

[0008] Among them, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, lichenification; the scoring information is one of the preset N2 scoring values; wherein N2 is an integer greater than or equal to 2, and the severity of the skin lesions in the lesion area corresponding to each of the N2 scoring values ​​is different.

[0009] In a second aspect, an embodiment of the present application provides an image processing device, the device comprising:

[0010] An acquisition unit, configured to acquire images of N1 parts of the target object; wherein N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image;

[0011] a segmentation unit, for performing image segmentation on an image of each of the N1 parts to obtain image data of a lesion region of the part; the lesion region is an idiopathic dermatitis region, and each of the N1 parts includes at least one lesion region;

[0012] A training unit, for training the image data of the lesion region of each of the N1 parts using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion region of the part;

[0013] Among them, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, lichenification; the scoring information is one of the preset N2 scoring values; wherein N2 is an integer greater than or equal to 2, and the severity of the skin lesions in the lesion area corresponding to each of the N2 scoring values ​​is different.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory and a processor, wherein the memory stores computer executable instructions, and when the processor runs the computer executable instructions on the memory, the image processing method described in the above embodiment can be implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium having executable instructions stored thereon, which, when executed by a processor, can implement the image processing method described in the above embodiment.

[0016] The technical solution of the embodiment of the present application can obtain the results of classification and identification of atopic dermatitis in various parts of the body of the target object by collecting images of various parts of the body of the target object, segmenting the lesion areas in the images of various parts of the body of the target object, and training the image data of the segmented lesion areas, thereby providing a reference for doctors to classify and identify AD in various parts of the body of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the image processing method provided in the embodiment of the present application Figure 1 ;

[0018] FIG2( a ) is a schematic diagram of a lesion image provided in an embodiment of the present application;

[0019] FIG2( b ) is a schematic diagram of a non-lesion image provided in an embodiment of the present application;

[0020] Figure 3 Schematic diagram 2 of the process of the image processing method provided in the embodiment of the present application;

[0021] Figure 4The skin segmentation and lesion segmentation results of the left front arm image of the upper limb provided in the embodiment of the present application;

[0022] Figure 5 The skin segmentation and lesion segmentation results of the image of the posterior part of the lower limb provided in the embodiment of the present application;

[0023] Figure 6 Schematic diagram of the image processing method provided in the embodiment of the present application Figure 3 ;

[0024] Figure 7 A schematic diagram of the composition of an image acquisition device provided in an embodiment of the present application;

[0025] Figure 8 A schematic diagram of the image processing process provided in an embodiment of the present application;

[0026] Fig. 9 A schematic diagram of the structure of an image processing device provided in an embodiment of the present application;

[0027] Fig.10 A schematic diagram of the hardware composition of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0029] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not necessarily need to be defined or explained in the subsequent drawings.

[0030] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0031] The severity of AD is assessed using the Eczema Area and Severity Index (EASI). The EASI score includes five items: affected area, erythema, infiltration or papules, exfoliation, and lichenification. The scores of the head and neck, upper limbs, trunk, and lower limbs are scored separately, and the total score is calculated by adding them up according to a specific formula.

[0032] If the EASI scoring method for evaluating AD relies on manual work, that is, it takes dozens of minutes for experienced clinicians to complete. The evaluation process is cumbersome and time-consuming, has a certain degree of subjectivity, and requires professional training to complete. It is difficult to popularize in doctors' daily clinical work. The technical solution of the embodiment of the present application uses artificial intelligence to assist doctors in classifying and identifying patients' AD. Only by uploading patient photos can AD of various parts of the patient's body be classified and identified, which is convenient for the subsequent calculation of the EASI of AD of the patient's whole body based on the AD identification results of various parts of the patient. The solution of the embodiment of the present application has at least the following effects:

[0033] (1) Provide more accurate and rapid identification and scoring methods for the assessment of patients' AD;

[0034] (2) Improve the work efficiency of clinicians, so that the standardized assessment of AD can be promoted and applied to primary care physicians and general practitioners, which is conducive to the promotion of chronic disease management of AD;

[0035] (3) Enable AD patients to complete self-monitoring of their condition and help reduce their physical and mental burden;

[0036] (4) Provide a comprehensive management basis for tertiary diagnosis and treatment of AD, thereby helping to reduce the socioeconomic burden.

[0037] The following is an introduction to the image processing method provided in the embodiments of the present application.

[0038] In the embodiments of the present application, the image processing method provided can be executed by a terminal device. The terminal device can be a device with data processing function, for example, a smart phone, a computer, a tablet computer and other smart devices. The terminal device can also be called a mobile terminal, a terminal, a mobile device, etc. The present application does not limit the type of the terminal device.

[0039] Figure 1 Schematic diagram of the image processing method provided in the embodiment of the present application Figure 1 ,like Figure 1 As shown, the image processing method provided in the embodiment of the present application includes the following steps:

[0040] S101: Acquire images of N1 parts of the target object.

[0041] In the embodiment of the present application, the target object is an AD patient, N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image.

[0042] In the embodiment of the present application, the N1 parts are one or more of the following parts: head and neck, upper limbs, trunk, and lower limbs.

[0043] In the embodiment of the present application, each of the N1 parts includes one or more sub-parts.

[0044] For example, the head and neck include sub-parts such as the face, head, front neck, and back neck; the upper limbs include sub-parts such as the front of the left arm, the back of the left arm, the front of the right arm, the back of the right arm, both arms, and the palms of both hands; the torso includes sub-parts such as the chest and abdomen, back, and perineum; the lower limbs include sub-parts such as the front of the thigh, the back of the thigh and buttocks, the front of the calf, the back of the calf, the base of both feet, and the backs of both feet.

[0045] In the embodiment of the present application, an image acquisition device can be used to acquire images of N1 parts of the whole body of the target object to obtain images of the N1 parts of the whole body of the target object. Specifically, when acquiring images, images of each sub-part of the N1 parts can be acquired respectively, and images of each sub-part of the N1 parts are acquired. The terminal device acquires the image of the N1 part of the target object by receiving the images of each sub-part of the N1 part of the target object acquired by the image acquisition device.

[0046] In the embodiment of the present application, the image acquisition device may be a part of the terminal device, or may be a module independently externally provided to the terminal device.

[0047] S102: For each of the N1 parts, perform image segmentation on the image of the part to obtain image data of the lesion area of ​​the part.

[0048] In the embodiment of the present application, the lesion area is an idiopathic dermatitis area, and each of the N1 parts includes at least one lesion area.

[0049] In an embodiment of the present application, when a terminal device acquires images of N1 parts of a target object, it performs image segmentation on each image included in the images of the N1 parts to obtain image data of the lesion area in each image of the N1 parts.

[0050] In the embodiment of the present application, for each image segmentation of the images of the N1 parts, a traditional image segmentation algorithm may be used for image segmentation, or a deep learning method may be used for image segmentation.

[0051] In one implementation, the above step S102 includes:

[0052] For each image of each of the N1 parts, a rejection model is used to determine whether the image contains a lesion area; if the rejection model determines that the image contains a lesion area, the image is segmented to obtain image data of the lesion area in the image; if the rejection model determines that there is no lesion area in the image, the image is discarded.

[0053] When the image acquisition device is performing image acquisition, some non-lesion photos may be captured due to misoperation. If these photos are directly transmitted to the terminal device, it will cause a waste of computing resources. The embodiment of the present application uses a rejection model to reject non-lesion photos and receive lesion photos.

[0054] The training data set of the rejection model training process consists of patient lesion images collected by the instrument (labeled as positive samples) and multiple (e.g., 1,000) images of other classes in the ILSVRC dataset (a dataset in the field of computer vision) as other images (labeled as negative samples). The training is based on EfficientNet (a convolutional neural network classifier). When the relevant numerical indicators such as loss (loss function) and acc (accuracy) reach the threshold, the training of the rejection model is completed.

[0055] After a captured image is passed to the rejection model, the rejection model judges the image. If it is determined that the image contains a lesion area, the image is transmitted to the terminal. If it is determined that the image does not contain a lesion area, the image is directly discarded to avoid wasting terminal computing resources.

[0056] As shown in Figure 2(a) and Figure 2(b), after the image shown in Figure 2(a) is transmitted to the rejection model, the rejection model determines that the image contains a lesion area, and then continues to perform image segmentation processing on the image to obtain the lesion area contained in the image; after the image shown in Figure 2(b) is transmitted to the rejection model, the rejection model determines that the image does not contain a lesion area, then discards the image and no longer performs image segmentation processing on the image, thereby reducing the waste of terminal computing resources.

[0057] S103: For each of the N1 parts, the image data of the lesion area of ​​the part is trained using a lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part.

[0058] In an embodiment of the present application, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, and lichenification; the scoring information is one of N2 preset scoring values; wherein N2 is an integer greater than or equal to 2, and each of the N2 scoring values ​​corresponds to a different severity of skin lesions in the lesion area.

[0059] In the embodiment of the present application, the image data of the lesion area includes color information of each pixel included in the lesion area; in one implementation manner, the above step S103 specifically includes the following steps:

[0060] Step 1-1): for each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain the confidence of at least one attribute information corresponding to each lesion area of ​​the part, and the confidence of the score value corresponding to each attribute information in the at least one attribute information;

[0061] Step 1-2): for each lesion area of ​​the part, the attribute information corresponding to the highest confidence value among the confidence values ​​of at least one attribute information corresponding to the lesion area is determined as the attribute information of the lesion area; and the score value corresponding to the attribute information of the lesion area with the highest confidence value is determined as the score information corresponding to the lesion area.

[0062] The terminal device receives images of each sub-part of the N1 parts of the target object collected by the image acquisition device. For each image of the N1 parts, the image data of the lesion area in each image is trained using a pre-trained lesion classification model to obtain an output vector, which is the erythema, papule, edema, epidermal exfoliation, lichenification and the confidence of their grading in the lesion area in each image, and each grading result with the highest confidence is selected as the classification result of the lesion area. For example, for the image of the left arm of the upper limb, the image of the part is identified by using the lesion classification model, and the lesion type corresponding to the lesion area of ​​the part image is erythema, papule, epidermal exfoliation, and lichenification, and the confidences are 0.7, 0.1, 0.1, and 0.2 respectively, then it is determined that the lesion type corresponding to the lesion area in the image of the part is erythema.

[0063] In an embodiment of the present application, if the severity score value is the preset N2 score values, which are specifically the following 5 score values: 0, 1, 1.5, 2, 2.5, 3, for the image of the left arm of the upper limb, the image of the part is identified by using the lesion classification model, and the lesion type corresponding to the lesion area of ​​the lesion image of the part is erythema. If the lesion type corresponding to the lesion area of ​​the lesion image of the part is erythema, the corresponding severity score values ​​are 0, 1, 1.5, 2, 2.5, 3, and the confidence levels are 0.1, 0.1, 0.1, 0.1, 0.6 respectively, then the model finally outputs the attribute information of the image of the left arm of the upper limb as erythema, and the specific severity score value (i.e., the score information) is 3.

[0064] The technical solution of the embodiment of the present application can obtain the results of classification and identification of atopic dermatitis in various parts of the body of the target object by collecting images of various parts of the body of the target object, segmenting the lesion areas in the images of various parts of the body of the target object, and training the image data of the segmented lesion areas, thereby providing a reference for doctors to classify and identify AD in various parts of the body of the target object.

[0065] Figure 3 Schematic diagram 2 of the flow chart of the image processing method provided in the embodiment of the present application; Figure 3 As shown, the image processing method provided in the embodiment of the present application includes the following steps:

[0066] S301: Acquire images of N1 parts of the target object.

[0067] The execution method of step S301 may refer to the above step S101, which will not be described in detail here.

[0068] S302: For each of the N1 parts, use the first image segmentation model to process the image of the part to obtain image data of the skin area in the image of the part.

[0069] S303: Using the second segmentation model, the image data of the skin area in the image of the part is processed to obtain the image data of the lesion area in the image of the part.

[0070] In an embodiment of the present application, in each image of each of the N1 parts, the image data of the skin area of ​​the image includes the number of pixels and the pixel size of the skin area, and the color information of each pixel, and the image data of the lesion area of ​​the image includes the number of pixels and the pixel size of the lesion area, and the color information of each pixel.

[0071] According to the EASI scoring rules, it is necessary to obtain the area of ​​the sampled skin and the area of ​​the lesion, and provide data support for the subsequent calculation of the EASI score based on the severity of the lesion and the ratio of the lesion to the total skin area.

[0072] The segmentation of the skin area in each image of each part can be implemented by using a first image segmentation model. Exemplarily, the first image segmentation model is a convolutional neural network model. The training of the first image segmentation model can be completed by a training data set including original images of skin data and polygon coordinates. After completing the training of the first image segmentation model, by inputting each image of each part into the first image segmentation model, the coordinates of the polygonal box corresponding to the skin area in the input image can be obtained; in addition, according to the part selected by the polygonal box, the area of ​​the skin area in the input image can be calculated by the size of the pixels included in the selected part and the pixel size.

[0073] Exemplarily, the first image segmentation model is used to perform skin segmentation on the image of the left arm front of the upper limb, and the following is obtained: Figure 4 Image data of a polygonal area of ​​skin area 1 is shown.

[0074] Exemplarily, the first image segmentation model is used to perform skin segmentation on the image of the lower leg, and the following is obtained: Figure 5 Image data of polygonal areas of skin area 2 and skin area 3 are shown.

[0075] According to the EASI scoring standard, each lesion in the acquired image needs to be segmented and classified in order to evaluate the occurrence of the lesion.

[0076] The segmentation of the lesion area in each image of each part can be implemented by using a second image segmentation model. For example, the first image segmentation model is a convolutional neural network model. The input skin area data is further segmented and the lesion area is extracted. The color information of the lesion can be obtained through the RGB channel information.

[0077] Exemplarily, the second image segmentation model is used to further identify the skin area 1 of the left arm in the upper limb part, and the following is obtained: Figure 4 Image data of lesion area 1 and lesion area 2 are shown.

[0078] Exemplarily, the second image segmentation model is used to further identify the skin area 2 and the skin area 3 at the back of the lower leg, and the following is obtained: Figure 5 Image data of lesion area 3 and lesion area 4 are shown.

[0079] S304: For each of the N1 parts, the image data of the lesion area of ​​the part is trained using a lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part.

[0080] For the lesion area in each image of each extracted part, the image data of the lesion area is trained using a lesion classification model to determine which type of lesion the lesion area is; and to determine the severity score value corresponding to the lesion type of the lesion area.

[0081] S305: For each of the N1 parts, calculate a first ratio corresponding to the part according to image data of the skin area in the image of the part, image data of the lesion area in the image of the part, and the ratio of the part to the whole body skin of the target object.

[0082] In the embodiment of the present application, the first ratio of each of the N1 parts is the ratio of the lesion area of ​​each part to the whole body skin area of ​​the target object.

[0083] S306: Accumulate the calculated first ratios corresponding to the various parts to obtain a second ratio; the second ratio is the ratio of the target object's whole body lesion area to the target object's whole body skin area.

[0084] S307: Compare the second ratio with a preset scoring table to obtain a first scoring value.

[0085] In the embodiment of the present application, the first score value is the score value of the whole body skin lesion area of ​​the target object. The preset score table is a correspondence table between the second ratio interval value and the preset score value.

[0086] For example, Figure 4 The image data of the polygonal area of ​​skin area 1 shown can be used to calculate the area of ​​skin area 1 included in the image of the left arm of the upper limb through the number of pixels and the size of the pixels included in the polygonal area of ​​skin area 1. The area of ​​the lesion area included in the image of the left arm of the upper limb can be calculated through the number of pixels and the size of the pixels included in lesion area 1 and lesion area 2.

[0087] For example, Figure 5The image data of the polygonal areas of skin area 2 and skin area 3 shown in the figure can be used to calculate the area of ​​skin area 2 and skin area 3 included in the image of the posterior part of the lower limb through the number of pixels and the size of the pixel included in the polygonal areas of skin area 2 and skin area 3. The area of ​​the lesion area included in the image of the posterior part of the lower limb can be calculated through the number of pixels and the size of the pixel included in the lesion area 3 and lesion area 4.

[0088] Table 1 below shows the proportion of each part of the human body to the surface area of ​​the human skin, and Table 2 is a preset scoring table for the skin lesion area of ​​the human body.

[0089] Table 1 The proportion of various parts of the human body to the surface area of ​​the human skin

[0090]

[0091]

[0092] Notes: 1. The hands occupy a total of 5% of the human body surface area (28%*18%=5%), of which the back of the palms accounts for 1% each and the skin between the hands accounts for 1%; 2. For children under 12 years old, the head and neck area is 9%+(12-age)%, the lower limb area is 46%-(12-age)%, and other parts are the same as adults.

[0093] Table 2 Preset scoring table for human body skin lesion area

[0094] Damage area 0 1-9％ 10-29％ 30-49％ 50-69％ 70-89％ 90-100％ Score 0 1 2 3 4 5 6

[0095] The calculation example is as follows:

[0096] For an 8-year-old patient, the total number of pixels of the lesion site in front of the unilateral calf was 4039, and the skin area in front of the unilateral calf was 10672 pixels. The number of pixels of the lesion site in the chest and abdomen was 16453 pixels, and the total number of pixels was 20231 pixels.

[0097] Then his calculation process of skin lesion area is:

[0098] 1. Determine the proportion of unilateral calf lesions to the whole body:

[0099] (4039 / 10672)*14%*(46%-(12-8)%) / 2=1.08%

[0100] 2. Determine the proportion of lesions in the trunk to the whole body:

[0101] 16453 / 20231*27%*48%=10.53%

[0102] 3. Add up the proportions of each part and enter it into the preset scoring table to obtain the skin lesion area score (i.e., the first scoring value): 1.08% + 10.53% > 10%. Referring to Table 2, the skin lesion area score of the 8-year-old patient is 2.

[0103] S308: Calculate a second score value according to the attribute information corresponding to each lesion area included in each of the N1 parts, the score information corresponding to each lesion area included in each of the N1 parts, and the weight coefficient corresponding to each part.

[0104] S309: Add the first score value and the second score value to obtain a third score value of systemic atopic dermatitis of the target object.

[0105] In the embodiment of the present application, the third scoring value is used to characterize the severity of the systemic atopic dermatitis of the target subject.

[0106] In one embodiment, the N1 parts are one or more of the following parts: head and neck, upper limbs, trunk, and lower limbs; each of the N1 parts includes one or more sub-parts; the age ranges corresponding to the target objects are different, and the weight coefficients corresponding to each of the N1 parts are different.

[0107] The following Table 3 is an EASI scoring table for target subjects aged over 8 years old, and Table 4 is an EASI scoring table for target subjects aged 8 years old or less.

[0108] Table 3 EASI score table for target subjects aged over 8 years old

[0109]

[0110] Table 4 EASI score table for target subjects aged 8 years or younger

[0111]

[0112] For example, a 22-year-old patient has a score of 2 for head and neck erythema, 1.5 for exfoliation of the trunk, 1 for lichenification of the upper limbs, and no abnormality in the lower limbs. The patient's total skin lesion area score (i.e., the first score) is 4. Then his EASI score (i.e., the third score of the patient's systemic atopic dermatitis) is:

[0113] 2*0.1+1.5*0.3+1*0.2+0*0.4+4=4.85.

[0114] Among them, the part 2*0.1+1.5*0.3+1*0.2+0*0.4=0.85 corresponds to the second score value.

[0115] The technical solution of the embodiment of the present application can obtain the results of classification and identification of atopic dermatitis in various parts of the body of the target object by collecting images of various parts of the body of the target object, segmenting the lesion areas in the images of various parts of the body of the target object, and training the image data of the segmented lesion areas, thereby providing a reference for doctors to classify and identify AD in various parts of the body of the target object.

[0116] Figure 6 Schematic diagram of the image processing method provided in the embodiment of the present application Figure 3 ;like Figure 6 As shown, the image processing method provided in the embodiment of the present application includes the following steps:

[0117] S601: Acquire images of N1 parts of the target object.

[0118] The implementation method of step S601 may refer to the above step S101 and will not be described in detail here.

[0119] S602: For each of the N1 parts, perform image segmentation on the image of the part to obtain image data of the lesion area of ​​the part.

[0120] The implementation method of step S602 may refer to the above steps S302 and S303, which will not be described again here.

[0121] S603: For each of the N1 parts, the image data of the lesion area of ​​the part and the additional information of the target object are trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part; the additional information includes at least one of the following: age, gender, and medical history.

[0122] For the lesion area in each image of each extracted part, the image data of the lesion area is trained using a lesion classification model to determine which type of lesion the lesion area is; and to determine the severity score value corresponding to the lesion type of the lesion area.

[0123] The technical solution of the embodiment of the present application is aimed at the characteristics of atopic dermatitis, and mainly considers the recognition of images of lesion areas in the collected pictures, while other factors such as patient gender, age, medical history information, etc. can be used as auxiliary information to obtain more accurate evaluation results.

[0124] The embodiment of the present application inputs the image data of the lesion area into the lesion classification model, and provides information such as the patient's gender, age, medical history, etc. After the model calculation, it outputs four major severity categories: erythema, papules / edema, epidermal exfoliation, and lichenification (for example, five severity levels of 0, 1, 1.5, 2, 2.5, and 3).

[0125] In one embodiment, the lesion classification model includes three parts: a first network, a second network, and a third network. The process of using the lesion classification model to train the image data of the lesion area of ​​the part and the additional information of the target object includes the following steps:

[0126] Step 6-1): for each of the N1 parts, inputting image data of the lesion area of ​​the part into the first network of the lesion classification model to obtain a first feature map; the first network includes at least one convolutional layer;

[0127] Step 6-2): inputting the first feature map into the second network of the lesion classification model to obtain N3 second feature maps; the second network includes multiple convolutional layers, and the N3 second feature maps are feature maps output by each convolutional layer in some convolutional layers of the multiple convolutional layers; wherein N3 is an integer greater than or equal to 1;

[0128] Step 6-3): Input the N3 second feature maps and the additional information into the third network of the lesion classification model to obtain the attribute information and scoring information corresponding to each lesion area of ​​the part.

[0129] In an embodiment of the present application, the first feature map includes low-level features of the image data of the lesion area, including global information of the image data, such as skin quality, lesion area size, lesion contour and other information; each of the N3 second feature maps includes high-level features of the image data of the lesion area, specifically semantic features of the image data of the lesion area.

[0130] The embodiment of the present application uses the first feature map and the second feature map to describe the image data from a global and partial perspective, thereby improving the effectiveness of the lesion classification model in judging the lesion type and the corresponding score.

[0131] Specifically, the data training process of the first network is as follows:

[0132] The input 224*224 three-channel RGB image (corresponding to the collected image of a certain part of the target object's body) is first passed into the convolution kernel size of 3*3 and the step size of 1. There are 64 convolution kernels in this layer for sampling. After two consecutive layers of the same convolution layer sampling, the output is a 224*224*64 feature map. After passing through the convolution layer of the previous layer, it is passed to the maximum pooling layer of the next layer for a 2*2 pooling operation, and the output is a 112*112*64 feature map. Similarly, the next convolution layer and pooling layer are similar to the previous step, except that the convolution kernel of conv-3 64 is replaced by the convolution kernel of conv-3 128. Correspondingly, the output of the convolution layer is 112*112*128, and the output of the pooling layer is 56*56*128. Furthermore, the input convolution kernel size is 3*3, the step size is 1, and there are 256 convolution kernels for sampling in this layer. This stage consists of 3 consecutive layers. After sampling in two identical convolution layers, a 56*56*256 feature map is obtained. Continuing, the output result of the previous convolution layer is passed to the maximum pooling layer to obtain a 28*28*256 feature map.

[0133] Similarly, the subsequent convolutional layer and pooling layer are similar to the above operations, except that the convolution kernel of conv-3 256 is replaced by the convolution kernel of conv-3 512. Correspondingly, the output of the convolutional layer is 56*56*512, and the output of the pooling layer is 28*28*512. Finally, the 28*28*512 is passed into the convolution kernel of conv-3 1024 with a step size of 2 to obtain a feature map of size 28*28*1024. At this point, the data training process of the first network is completed.

[0134] The data training process of the second network is as follows:

[0135] For the feature map output by the first network, it is passed to the convolution layer composed of conv-3 1024 and convolution kernel with a step size of 1 for processing, and a 28*28*1024 feature map is obtained. This feature map is passed to the next layer and is also input as part of the fully connected layer for processing in the third part of the network. The result of the previous layer is continuously convolved to the next layer to further obtain subsequent features. After passing through the convolution layer with conv-3 512, a step size of 1 and a step size of 2, a 28*28*512 feature map is obtained. It is passed to the subsequent convolution layer and is also processed as part of the third part of the network. Further, after passing through the convolution layer with conv-3 256, a step size of 1 and a step size of 2, a 28*28*256 feature map is obtained. It is passed to the subsequent convolution layer and is also processed as part of the third part of the network. Finally, after conv-3 128, the convolution with step size 1 and step size 2 respectively, the difference from the previous one is that the two 28*28*128 feature maps obtained at this time are used as the input values ​​of the third part of the network. At this point, the data training process of the second network is completed.

[0136] The data training process of the third network is as follows:

[0137] The feature map obtained by the previous second network is combined with the vector encoding of the target object's medical history information, gender, age and other factors to form a new vector. After global average pooling, it is passed to the fully connected layer. After activation by the sigmoid function, an output vector can be obtained, which is the confidence of erythema, papule, edema, epidermal exfoliation, lichenification and their grading. The grading result with the highest confidence is selected as the prediction result of the lesion condition.

[0138] In an embodiment of the present application, the first network and the second network of the lesion classification model process the image data, and the third network combines other additional information and the feature map to form a new vector, so that the weight of the auxiliary information is smaller, which is beneficial to improving the accuracy of subsequent AD scoring.

[0139] In one embodiment, the initial lesion classification model can be trained using training data to obtain a lesion classification model; wherein the training data includes at least one of the following information of N4 lesion images: labeled lesion classification information, size information of each pixel, color information of each pixel, and evaluation information; the labeled lesion classification information includes: evaluation results, gender, age, and medical history; wherein N4 is an integer greater than or equal to 1.

[0140] Specifically, the training of the lesion classification model requires the following steps:

[0141] a. Obtain training data, which includes the lesion classification information annotated by experts, the lesion pixel size and RGB information calculated above, and the doctor's evaluation information. The information to be annotated includes the expert evaluation results, the patient's gender, age, medical history, etc.

[0142] b. Quantify the evaluation information and perform quantitative encoding on the diagnostic corpus. The specific method is to use word vector encoding to map the expert advice corpus into vectors.

[0143] Here, the evaluation information may specifically be diagnostic suggestion corpus related to atopic dermatitis, and the evaluation result may specifically be a diagnostic result for atopic dermatitis.

[0144] An embodiment of the present application provides a lesion classification model. During the training process of the lesion classification model on the image data of the lesion area, the body part of the target object is used as input in combination with relevant information such as the patient's gender, age, genetic history, etc., and the comprehensively calculated EASI score is used as output.

[0145] The lesion classification model of the embodiment of the present application can extract features from the RGB image of the atopic dermatitis lesions based on the color, shape and other characteristics of the lesions, but not only considers the feature information in the image, but also considers gender, age and medical history as auxiliary diagnosis factors, and integrates the image and the relevant auxiliary diagnosis data of the target object, which is added as training data to the training of the model, thereby improving the accuracy of lesion classification and identification.

[0146] Next, an image acquisition device for acquiring images of N1 parts of a target object is introduced.

[0147] like Figure 7 As shown, the image acquisition device includes: a solid color background 701, a positioning color block 702, a longitudinal slide rail 703, a transverse slide rail 704, a camera module 705, a retractable rotating device 706, a colored laser 707, a shadowless lamp 708, a communication module 709, and a host 710.

[0148] The image capture process is as follows:

[0149] In the first step, the target object stands at the designated position of the image acquisition device, and the ultra-wide-angle camera in the camera module 705 takes a full-body photo of the human body. The photo should include the entire human body and a positioning color block 702 of a solid color background 701. A coordinate system is established based on the position of the positioning color block 702 to facilitate the positioning of various parts of the human body;

[0150] The second step is to extract the key points of the human body according to the captured full-body image, determine the position of each part of the human body, and at the same time, obtain the relative position and corresponding position coordinates of each part according to the positioning color block 702;

[0151] The third step is to control the camera module 705 to the position above each part of the human body according to the obtained position coordinate information, and use the laser position determination algorithm and the laser position adjustment device (including the longitudinal slide rail 703, the transverse slide rail 704, and the retractable rotation device 706) to adjust the colored laser 707 to irradiate the background next to the human body, control the camera module 705 to capture images of each part in turn, and use the communication module 709 to transmit the captured images to the host 710.

[0152] The above image acquisition device and the process of using the image acquisition device to acquire images of various parts of the body of the target object are merely an exemplary implementation. The embodiment of the present application does not impose specific restrictions on the image acquisition device and the process of using the image acquisition device to acquire images of various parts of the body of the target object, as long as images of various parts of the human body can be acquired.

[0153] Figure 8 A schematic diagram of the image processing process provided in the embodiment of the present application; Figure 8 As shown in Figure 1, the lesion classification process includes the following steps:

[0154] S801: Collect images.

[0155] An image acquisition device is used to acquire images of various parts of the patient's body, specifically including images of various sub-parts contained in the patient's head and neck, upper limbs, torso, and lower limbs. Among them, the head and neck include sub-parts such as the face, head, front neck, and back neck; the upper limbs include sub-parts such as the front of the left arm, the back of the left arm, the front of the right arm, the back of the right arm, both arms, and the palms of both hands; the torso includes sub-parts such as the chest and abdomen, the back, and the perineum; the lower limbs include sub-parts such as the front of the thigh, the back of the thigh and buttocks, the front of the calf, the back of the calf, the base of both feet, and the back of both feet.

[0156] S802: Reject model judgment.

[0157] When the image acquisition device is performing image acquisition, it may capture some non-lesion photos due to misoperation. If these photos are directly transmitted to the terminal device, it will cause a waste of computing resources. The rejection model is used to reject non-lesion photos, receive lesion photos, and perform subsequent skin segmentation and lesion segmentation steps.

[0158] S803: Skin segmentation.

[0159] The segmentation of the skin area in each image of each part can be achieved by using the first image segmentation model. By inputting each image of each part into the first image segmentation model, image data corresponding to the skin area in the input image can be obtained.

[0160] S804: Lesion segmentation.

[0161] The segmentation of the lesion area in each image of each part can be achieved by using a second image segmentation model, and the input skin area data is further segmented and the corresponding image data of the lesion area is extracted. The color information of the lesion can be obtained through the RGB channel information.

[0162] S805: Lesion classification.

[0163] For the lesion area in each image of each extracted part, the image data of the lesion area is trained using a lesion classification model to determine which type of lesion the lesion area is; and to determine the severity score value corresponding to the lesion type of the lesion area.

[0164] S806: Calculate the EASI score.

[0165] The method of calculating the EASI score value may refer to the above steps S305 to S309, which will not be described in detail here.

[0166] Fig. 9 A schematic diagram of the structure of an image processing device 900 provided in an embodiment of the present application is shown in FIG. Fig. 9 As shown, the lesion classification device 900 includes:

[0167] An acquisition unit 901 is used to acquire images of N1 parts of a target object; wherein N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image;

[0168] The segmentation unit 902 is used to perform image segmentation on the image of each of the N1 parts to obtain image data of a lesion area of ​​the part; the lesion area is an idiopathic dermatitis area, and each of the N1 parts includes at least one lesion area;

[0169] A training unit 903 is used to train the image data of the lesion area of ​​each of the N1 parts using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part;

[0170] Among them, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, lichenification; the scoring information is one of the preset N2 scoring values; wherein N2 is an integer greater than or equal to 2, and the severity of the skin lesions in the lesion area corresponding to each of the N2 scoring values ​​is different.

[0171] In one embodiment, the segmentation unit is used to: for each of the N1 parts, use the first image segmentation model to process the image of the part to obtain image data of the skin area in the image of the part; use the second segmentation model to process the image data of the skin area in the image of the part to obtain image data of the lesion area in the image of the part.

[0172] In one embodiment, the image data of the skin area includes the number of pixels and the pixel size of the skin area, the image data of the lesion area includes the number of pixels and the pixel size of the lesion area, and the device further includes:

[0173] A first calculation unit is used to calculate, for each of the N1 parts, a first ratio corresponding to the part according to image data of the skin area in the image of the part, image data of the lesion area in the image of the part, and the ratio of the part to the whole body skin of the target object; the first ratio of the part is the ratio of the lesion area of ​​the part to the whole body skin area of ​​the target object;

[0174] A first accumulation unit is used to accumulate the calculated first ratios corresponding to the various parts to obtain a second ratio; the second ratio is the ratio of the whole body lesion area of ​​the target object to the whole body skin area of ​​the target object;

[0175] A comparison unit is used to compare the second ratio with a preset scoring table to obtain a first scoring value; the first scoring value is a scoring value of the whole body skin lesion area of ​​the target object; the preset scoring table is a correspondence table between the second ratio interval value and the preset scoring value.

[0176] In one embodiment, the device further comprises:

[0177] A second calculation unit is used to calculate a second score value according to the attribute information corresponding to each lesion area included in each of the N1 parts, the score information corresponding to each lesion area included in each of the N1 parts, and the weight coefficient corresponding to each part;

[0178] The second accumulating unit is used to add the first scoring value and the second scoring value to obtain a third scoring value of the systemic atopic dermatitis of the target object; the third scoring value is used to characterize the severity of the systemic atopic dermatitis of the target object.

[0179] In one embodiment, the N1 parts are one or more of the following parts: head and neck, upper limbs, trunk, and lower limbs; each of the N1 parts includes one or more sub-parts; the age ranges corresponding to the target objects are different, and the weight coefficients corresponding to each of the N1 parts are different.

[0180] In one embodiment, the segmentation unit is used to use a rejection model to determine, for each image of each of the N1 parts, whether the image contains a lesion area; if the rejection model determines that the image contains a lesion area, the image is segmented to obtain image data of the lesion area in the image; if the rejection model determines that there is no lesion area in the image, the image is discarded.

[0181] In one embodiment, the training unit is used to train the image data of the lesion area of ​​each of the N1 parts and the additional information of the target object using a lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part; the additional information includes at least one of the following: age, gender, and medical history.

[0182] In one embodiment, the training unit is used to input the image data of the lesion area of ​​each of the N1 parts into the first network of the lesion classification model to obtain a first feature map; the first network includes at least one convolutional layer; the first feature map includes low-level features of the image data of the lesion area of ​​the part; the first feature map is input into the second network of the lesion classification model to obtain N3 second feature maps; the second network includes multiple convolutional layers, and the N3 second feature maps are feature maps output by each convolutional layer in some of the multiple convolutional layers; each of the N3 second feature maps includes high-level features of the image data of the lesion area of ​​the part; wherein N3 is an integer greater than or equal to 1; the N3 second feature maps and the additional information are input into the third network of the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part.

[0183] In one embodiment, the training unit is used to train the initial lesion classification model using training data to obtain a lesion classification model; wherein the training data includes at least one of the following information of N4 lesion images: labeled lesion classification information, size information of each pixel, color information of each pixel, and evaluation information; the labeled lesion classification information includes: evaluation results, gender, age, and medical history; wherein N4 is an integer greater than or equal to 1.

[0184] In one embodiment, the image data of the lesion area includes color information of each pixel included in the lesion area; the training unit is used to train the image data of the lesion area of ​​each of the N1 parts using a lesion classification model to obtain the confidence of at least one attribute information corresponding to each lesion area of ​​the part, and the confidence of the score value corresponding to each attribute information in the at least one attribute information; for each lesion area of ​​the part, the attribute information corresponding to the highest confidence value among the confidence values ​​of at least one attribute information corresponding to the lesion area is determined as the attribute information of the lesion area; and the score value with the highest confidence value corresponding to the attribute information of the lesion area is determined as the score information corresponding to the lesion area.

[0185] Those skilled in the art should understand that Fig. 9 The functions implemented by each unit in the image processing device shown can be understood by referring to the relevant description of the aforementioned image processing method. Fig. 9 The functions of each unit in the image processing device shown may be implemented by a program running on a processor, or may be implemented by a specific logic circuit.

[0186] An embodiment of the present application also provides an electronic device. Fig.10 Schematic diagram of the hardware structure of the electronic device of the embodiment of the present application, such as Fig.10 As shown, the electronic device includes: a communication component 1003 for data transmission, at least one processor 1001, and a memory 1002 for storing a computer program that can be run on the processor 1001. The various components in the terminal are coupled together through a bus system 1004. It can be understood that the bus system 1004 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 1004 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, Fig.10 Various buses are labeled as bus system 1004 .

[0187] When the processor 1001 executes the computer program, at least Figure 1 , Figure 3 or Figure 6 The steps of the method are shown.

[0188] It can be understood that the memory 1002 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disk, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), and direct RAM bus random access memory (DRRAM, Direct Rambus Random Access Memory).The memory 1002 described in the embodiments of the present application is intended to include but is not limited to these and any other suitable types of memories.

[0189] The method disclosed in the above embodiment of the present application can be applied to the processor 1001, or implemented by the processor 1001. The processor 1001 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of the hardware in the processor 1001 or an instruction in the form of software. The above processor 1001 may be a general-purpose processor, a DSP, or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 1001 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or any conventional processor, etc. In combination with the steps of the method disclosed in the embodiment of the present application, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in the memory 1002, and the processor 1001 reads the information in the memory 1002 and completes the steps of the above method in combination with its hardware.

[0190] In an exemplary embodiment, the electronic device may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD), FPGA, general-purpose processor, controller, MCU, microprocessor, or other electronic components to execute the aforementioned method for identifying atopic dermatitis.

[0191] The present application also provides a computer-readable storage medium having a computer program stored thereon, wherein the program is at least used to execute when executed by a processor Figure 1 , Figure 3 or Figure 6 The computer readable storage medium may be a memory. The memory may be Fig.10 Memory 1002 is shown.

[0192] The technical solutions described in the embodiments of the present application can be combined arbitrarily without conflict.

[0193] In the several embodiments provided in the present application, it should be understood that the disclosed methods and intelligent devices can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0194] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0195] In addition, all functional units in the embodiments of the present application may be integrated into a second processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0196] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An image processing method, characterized in that: The method comprises: Acquire images of N1 parts of the target object; wherein N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image; For each of the N1 parts, image segmentation is performed on the image of the part to obtain image data of the lesion area of ​​the part; the lesion area is an idiopathic dermatitis area, and each of the N1 parts includes at least one lesion area; For each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part; Among them, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, lichenification; the scoring information is one of the preset N2 scoring values; N2 is an integer greater than or equal to 2, and the severity of the skin lesions in the lesion area corresponding to each of the N2 scoring values ​​is different.

2. The method according to claim 1, characterized in that The step of performing image segmentation on the image of each of the N1 parts to obtain image data of the lesion area of ​​the part includes: For each of the N1 parts, use the first image segmentation model to process the image of the part to obtain image data of the skin area in the image of the part; The image data of the skin area in the image of the part is processed using the second segmentation model to obtain the image data of the lesion area in the image of the part.

3. The method according to claim 2, characterized in that The image data of the skin area includes the number of pixels and the pixel size of the skin area, the image data of the lesion area includes the number of pixels and the pixel size of the lesion area, and the method further includes: For each of the N1 parts, a first ratio corresponding to the part is calculated according to image data of a skin area in an image of the part, image data of a lesion area in an image of the part, and a ratio of the part to the whole body skin of the target object; the first ratio of the part is the ratio of the lesion area of ​​the part to the whole body skin area of ​​the target object; The calculated first ratios corresponding to the various parts are accumulated to obtain a second ratio; the second ratio is the ratio of the target object's whole body lesion area to the target object's whole body skin area; The second ratio is compared with a preset scoring table to obtain a first scoring value; the first scoring value is a scoring value of the whole body skin lesion area of ​​the target object; the preset scoring table is a correspondence table between the second ratio interval value and the preset scoring value.

4. The method according to claim 3, characterized in that The method further comprises: Calculating a second scoring value according to the attribute information corresponding to each lesion area included in each of the N1 parts, the scoring information corresponding to each lesion area included in each of the N1 parts, and the weight coefficient corresponding to each part; The first scoring value and the second scoring value are added to obtain a third scoring value of the systemic atopic dermatitis of the target object; the third scoring value is used to characterize the severity of the systemic atopic dermatitis of the target object.

5. The method according to claim 4, characterized in that The N1 parts are one or more of the following parts: head and neck, upper limbs, trunk, and lower limbs; each of the N1 parts includes one or more sub-parts; the age ranges corresponding to the target objects are different, and the weight coefficients corresponding to each of the N1 parts are different.

6. The method according to any one of claims 1 to 5, characterized in that The step of performing image segmentation on the image of each of the N1 parts to obtain image data of the lesion area of ​​the part includes: For each image of each of the N1 parts, using a rejection model to determine whether the image contains a lesion area; If the rejection model determines that the image contains a lesion area, the image is segmented to obtain image data of the lesion area in the image; if the rejection model determines that the image does not contain a lesion area, the image is discarded.

7. The method according to any one of claims 1 to 5, characterized in that For each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part, including: For each of the N1 parts, the lesion classification model is used to train the image data of the lesion area of ​​the part and the additional information of the target object to obtain the attribute information and scoring information corresponding to each lesion area of ​​the part; the additional information includes at least one of the following: age, gender, and medical history.

8. The method according to claim 7, characterized in that For each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part, including: For each of the N1 parts, inputting image data of the lesion area of ​​the part into the first network of the lesion classification model to obtain a first feature map; the first network includes at least one convolutional layer; the first feature map includes low-level features of the image data of the lesion area of ​​the part; Inputting the first feature map into the second network of the lesion classification model to obtain N3 second feature maps; the second network includes multiple convolutional layers, and the N3 second feature maps are feature maps output by each convolutional layer in some convolutional layers of the multiple convolutional layers; each of the N3 second feature maps includes high-level features of image data of the lesion area of ​​the part; wherein N3 is an integer greater than or equal to 1; The N3 second feature maps and the additional information are input into the third network of the lesion classification model to obtain the attribute information and scoring information corresponding to each lesion area of ​​the part.

9. The method according to claim 7, characterized in that: Before training the image data of the lesion area of ​​each of the N1 parts using the lesion classification model to obtain the attribute information and scoring information corresponding to each lesion area of ​​the part, the method further includes: The initial lesion classification model is trained using the training data to obtain a lesion classification model; wherein the training data includes at least one of the following information of N4 lesion images: labeled lesion classification information, size information of each pixel, color information of each pixel, and evaluation information; the labeled lesion classification information includes: evaluation results, gender, age, and medical history; wherein N4 is an integer greater than or equal to 1.

10. The method according to any one of claims 1 to 5, characterized in that The image data of the lesion area includes color information of each pixel included in the lesion area; for each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion area of ​​the part, including: For each of the N1 parts, the image data of the lesion area of ​​the part is trained using the lesion classification model to obtain the confidence of at least one attribute information corresponding to each lesion area of ​​the part, and the confidence of the score value corresponding to each attribute information of the at least one attribute information; For each lesion area of ​​the part, the attribute information corresponding to the highest confidence value among the confidence values ​​of at least one attribute information corresponding to the lesion area is determined as the attribute information of the lesion area; and the score value with the highest confidence value corresponding to the attribute information of the lesion area is determined as the score information corresponding to the lesion area.

11. An image processing device, characterized in that: The device comprises: An acquisition unit, configured to acquire images of N1 parts of the target object; wherein N1 is an integer greater than or equal to 1, and each of the N1 parts includes at least one image; a segmentation unit, for performing image segmentation on an image of each of the N1 parts to obtain image data of a lesion region of the part; the lesion region is an idiopathic dermatitis region, and each of the N1 parts includes at least one lesion region; A training unit, for training the image data of the lesion region of each of the N1 parts using the lesion classification model to obtain attribute information and scoring information corresponding to each lesion region of the part; Among them, the attribute information includes at least one of the following: erythema, papule, epidermal exfoliation, lichenification; the scoring information is one of the preset N2 scoring values; wherein N2 is an integer greater than or equal to 2, and the severity of the skin lesions in the lesion area corresponding to each of the N2 scoring values ​​is different.

12. An electronic device, characterized in that: The electronic device comprises: a memory and a processor, wherein the memory stores computer executable instructions, and the processor can implement any one of claims 1 to 10 when executing the computer executable instructions on the memory.

13. A computer storage medium, characterized in that: The storage medium stores executable instructions, which, when executed by a processor, implement the method according to any one of claims 1 to 10.

Images