Deep learning-based dermatosis auxiliary diagnosis method and system, and electronic equipment

Abstract

The invention relates to the technical field of deep learning medical image classification algorithms, in particular to a skin disease auxiliary diagnosis method and system based on deep learning and electronic equipment. The method comprises the following steps: acquiring two pictures containing body parts shot by a user in two periods, recording the two pictures as a first image and a second image, and respectively inputting the first image and the second image into a skin disease classification neural network model and a body part classification neural network model; identifying body parts of the first image and the second image, lesion areas of the body parts and skin disease types corresponding to the lesion areas, and calculating first areas of the lesion areas and second areas of the body parts where the lesion areas are located, according to the invention, synchronous intelligent judgment on whether various skin diseases suffered by the user are good or not and whether the skin diseases are transferred or not is realized, the diagnosis workload and the diagnosis precision of a doctor are improved to a great extent, the diagnosis time of the doctor is saved to a great extent, the diagnosis efficiency is improved, and the intelligent degree and the usability of the system are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning medical image classification algorithms, and in particular to a skin disease auxiliary diagnosis method and system, and electronic equipment based on deep learning. Background Art

[0002] The existing classification diagnosis of skin diseases based on deep learning is mainly to diagnose whether there are skin lesions, that is, to determine whether there are skin lesions in the acquired image. For example, the Chinese invention patent with the announcement number CN107909566A discloses "an image recognition method for skin cancer melanoma based on deep learning", and the method uses deep learning technology to diagnose whether there is melanoma. This method is too simple to only determine whether there are skin lesions. Whether there are lesions on the skin can be observed immediately by the naked eye of ordinary people, and when the skin has lesions and is not melanoma, the method may not be able to make a correct diagnosis, and it is even more impossible to determine whether the skin lesions have improved or metastasized. In addition, in the prior art, the same patient may suffer from multiple skin diseases, which further increases the difficulty of diagnosis. If the various skin diseases suffered by the patient are judged by visual observation, it is usually difficult, with large errors, and a large workload, which easily takes a lot of time for doctors to diagnose. In the prior art, there is a lack of intelligent judgment on whether the skin lesions have improved and whether they have metastasized.

[0003] Therefore, the prior art needs to be further developed. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned technical deficiencies and provide a skin disease auxiliary diagnosis method and system, and electronic equipment based on deep learning to solve the problems existing in the prior art.

[0005] To achieve the above technical objectives, according to a first aspect of the present invention, the present invention provides a method for auxiliary diagnosis of skin diseases based on deep learning, the method comprising:

[0006] S100, obtaining different skin lesion images, scaling the skin lesion images of different sizes to a preset size; performing data amplification on the scaled images according to a preset method; and training a neural network model for skin disease classification using the amplified images; obtaining different body part images, scaling the body part images of different sizes to a preset size; performing data amplification on the scaled images according to a preset method; and training a neural network model for body part classification using the amplified images;

[0007] S200, obtaining two pictures including body parts of the user taken by the user at two periods, respectively recorded as a first image and a second image, inputting the first image and the second image into a skin disease classification neural network model and a body part classification neural network model respectively, identifying each body part and the lesion area of ​​each body part in the first image and the second image, and the type of skin disease corresponding to the lesion area, calculating a first area of ​​each lesion area and a second area of ​​the body part where the lesion area is located, calculating each first ratio of the first area to the second area of ​​each body part in the first image, and calculating each second ratio of the first area to the second area of ​​each body part in the second image;

[0008] S300: Determine whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios.

[0009] Specifically, the body parts include:

[0010] Face, chest, arms, legs, abdomen, back.

[0011] Specifically, judging whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios includes:

[0012] Randomly select a skin disease suffered by the user, calculate the sum of the first ratios corresponding to the skin disease, and record it as the total proportion of the first lesion area of ​​the skin disease; and calculate the sum of the second ratios corresponding to the skin disease, and record it as the total proportion of the second lesion area of ​​the skin disease;

[0013] Determining whether the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease;

[0014] Determine whether the condition of the skin disease has improved based on the judgment result;

[0015] And calculate whether the condition of other types of skin diseases suffered by the user has improved according to the above method.

[0016] Specifically, judging whether the condition of the skin disease has improved according to the judgment result includes:

[0017] If the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease, it is determined that the skin disease has not improved, a prompt signal indicating that the skin disease has not improved is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​the skin disease are output;

[0018] If the total proportion of the second lesion area of ​​this type of skin disease is less than or equal to the total proportion of the first lesion area of ​​this type of skin disease, it is judged that this type of skin disease has improved, and a prompt signal regarding the improvement of this type of skin is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​this type of skin disease are output.

[0019] Specifically, the data amplification of the scaled image according to a preset method includes:

[0020] The scaled image is transformed by random cropping, random flipping, random rotation, and random deformation.

[0021] Specifically, the method further includes:

[0022] Whether the skin diseases suffered by the user have metastasized is determined according to the first ratios and the second ratios.

[0023] Specifically, judging whether the skin diseases suffered by the user have metastasized according to the first ratios and the second ratios includes:

[0024] Randomly select a skin disease suffered by the user, calculate the difference between the second ratio and the first ratio of each body part corresponding to the skin disease, respectively determine whether the difference is greater than a first preset threshold, and determine whether the skin disease suffered by the user has metastasized to each body part according to the determination result;

[0025] Calculate whether the user's other types of skin diseases have been transferred according to the above method.

[0026] Specifically, judging whether the skin disease of the user has metastasized to various parts of the body according to the judgment result includes:

[0027] If the difference of a certain body part is greater than a first preset threshold, it is determined that the skin disease has metastasized to the body part, a prompt signal indicating that the skin disease has metastasized to the body part is output, and the name of the body part is output;

[0028] If the difference of a certain body part is less than or equal to the first preset threshold, it is determined that the type of skin disease has not metastasized to the body part, a prompt signal is output indicating that the type of skin disease has not metastasized to the body part, and the name of the body part is output.

[0029] According to a second aspect of the present invention, a skin disease auxiliary diagnosis system based on deep learning is provided, comprising:

[0030] An acquisition module, used to acquire different skin lesion images; or used to acquire different body part images; or used to acquire two images of the user's body parts taken by the user at two periods;

[0031] A control module is used to scale skin lesion images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a skin disease classification neural network model; or to scale body part images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a body part classification neural network model; or to record two images of the user's body parts taken by the user at two periods as a first image and a second image, respectively, and input the first image and the second image into a skin disease classification neural network model and a body part classification neural network model, respectively, identify each body part and the lesion area of ​​each body part in the first image and the second image, and the skin disease type corresponding to the lesion area, calculate the first area of ​​each lesion area and the second area of ​​the body part where the lesion area is located, calculate each first ratio of the first area to the second area of ​​each body part in the first image, and calculate each second ratio of the first area to the second area of ​​each body part in the second image; or to judge whether the condition of each skin disease suffered by the user has improved according to each first ratio and each second ratio.

[0032] According to a third aspect of the present invention, there is provided an electronic device, comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the above-mentioned deep learning-based auxiliary diagnosis method for skin diseases is implemented.

[0033] Beneficial effects:

[0034] The present invention can realize synchronous intelligent judgment on whether various skin diseases suffered by the user have improved and whether metastasis has occurred by only obtaining two pictures containing the user's body parts taken by the user at two time periods. It greatly reduces the doctor's diagnostic workload and diagnostic accuracy, saves the doctor's diagnostic time to a great extent, improves the diagnostic efficiency, greatly improves the intelligence and usability of the present invention, and greatly expands the application scenarios of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flow chart of a method for auxiliary diagnosis of skin diseases based on deep learning provided in a specific embodiment of the present invention;

[0036] Figure 2 It is a schematic diagram of the system composition of a skin disease auxiliary diagnosis system based on deep learning provided in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is clearly and completely described below in conjunction with the accompanying drawings of the present invention. Based on the embodiments in this application, other similar embodiments obtained by ordinary technicians in this field without making creative work should all fall within the scope of protection of this application. In addition, the directional words mentioned in the following embodiments, such as "up", "down", "left", "right", etc., are only reference to the directions of the accompanying drawings. Therefore, the directional words used are used to illustrate rather than limit the invention.

[0038] The present invention will be further described below in conjunction with the accompanying drawings and preferred embodiments.

[0039] See also Figure 1 The present invention provides a skin disease auxiliary diagnosis method based on deep learning, comprising:

[0040] S100, obtaining different skin lesion pictures, scaling the skin lesion pictures of different sizes to a preset size; performing data amplification on the scaled pictures according to a preset method; using the amplified pictures to train a neural network model for skin disease classification; obtaining different body part pictures, scaling the body part pictures of different sizes to a preset size; performing data amplification on the scaled pictures according to a preset method; using the amplified pictures to train a neural network model for body part classification.

[0041] It should be noted that the steps before step S100 include:

[0042] A first preset threshold and a preset size are preset.

[0043] It can be understood that the first preset threshold and the preset size can be specifically set according to actual conditions, and the present invention does not limit the specific values ​​of the first preset threshold and the preset size.

[0044] Preferably, the present invention sets the first preset threshold to 3%. The first preset threshold is set to 3% based on the conclusion of the technical personnel of the present application through a large number of experiments, which can better distinguish whether the skin disease has metastasized, and further improves the reliability and usability of the present invention.

[0045] Preferably, the present invention sets the preset size to a resolution of 331*331.

[0046] Specifically, the data amplification of the scaled image according to a preset method includes:

[0047] The scaled image is subjected to transformations including random cropping, random flipping, random rotation and random deformation. All transformations have equal probability of occurring, that is, the transformation may or may not occur.

[0048] Specifically, the body parts include:

[0049] Face, chest, arms, legs, abdomen, back.

[0050] It should be noted here that the use of the amplified images to train the skin disease classification neural network model includes:

[0051] S1. Use the amplified images to construct a data set, and divide the constructed data set into a training set, a test set, and a validation set;

[0052] S2. Use the k-means clustering algorithm to cluster the images in the data set and obtain K prior boxes;

[0053] S3. Add 1×1 convolution kernels before the first and second 3×3 convolution layers of Backbone of the YOLO v5 algorithm; Add Non-loca l attention mechanism modules before the third and fourth Bott l eneck CSP of Backbone of the YOLO v5 algorithm;

[0054] S4. Use the CSPNet structure composed of Bottleneck and standard convolutional layers in the Backbone of the YOLO v5 algorithm and add the Output detection layer; reference the PANet structure in the Neck of the YOLO v5 algorithm; generate the improved YOLO v5 algorithm;

[0055] S5. Train and test the improved YOLO v5 algorithm on the established skin disease dataset.

[0056] Furthermore, in step S1, skin disease pictures are collected and the obtained picture format is converted into .jpg format. The pictures are labeled with types using label Img software, and the output format is .xml file, which is made into a standard VOC data set format. The constructed data set is then randomly divided into training set, test set and validation set in an 8:1:1 ratio, and each skin disease is labeled with an independent skin disease name label to complete the construction of the skin disease data set.

[0057] Further, in step S5, the training and testing on the skin disease dataset specifically includes:

[0058] (1) Set the parameters of the improved YOLO v5 algorithm: the training step size is set to 300, the initial learning rate is set to 0.01, and the batch_size is set to 1;

[0059] (2) Use the k-means clustering algorithm to obtain K target anchor values ​​and modify the original anchor box to the new anchor box after clustering;

[0060] (3) The input image is first subjected to Mosaic data enhancement and image normalization, and the image used for training is normalized to a size of 640×640;

[0061] (4) The Backbone part slices and convolves the image to obtain a feature map;

[0062] (5) The feature information of the feature map is extracted through the Neck part. In the Neck part, the image is first upsampled to extract image features, and then downsampled to fuse feature information of different scales. Finally, the detection result is output from the output end to obtain the detection model;

[0063] (6) Use the validation set to verify the performance of the model.

[0064] It should be noted that the method of using the amplified images to train the body part classification neural network model includes:

[0065] S1. Use the amplified images to construct a data set, and divide the constructed data set into a training set, a test set, and a validation set;

[0066] S2. Use the k-means clustering algorithm to cluster the images in the data set and obtain K prior boxes;

[0067] S3. Add 1×1 convolution kernels before the first and second 3×3 convolution layers of Backbone of the YOLO v5 algorithm; Add Non-loca l attention mechanism modules before the third and fourth Bott l eneck CSP of Backbone of the YOLO v5 algorithm;

[0068] S4. Use the CSPNet structure composed of Bottleneck and standard convolutional layers in the Backbone of the YOLO v5 algorithm and add the Output detection layer; reference the PANet structure in the Neck of the YOLO v5 algorithm; generate the improved YOLO v5 algorithm;

[0069] S5. Train and test the improved YOLO v5 algorithm on the established body part dataset.

[0070] Furthermore, in step S1, body part pictures are collected and the obtained picture format is converted into .jpg format, and the picture types are annotated using labe l Img software. The output format is .xml file, which is made into a standard VOC data set format. The constructed data set is then randomly divided into a training set, a test set, and a validation set in a ratio of 8:1:1, and each body part is labeled with an independent body part name label to complete the construction of the body part data set.

[0071] Furthermore, in step S5, the training and testing on the body part data set specifically includes:

[0072] (1) Set the parameters of the improved YOLO v5 algorithm: the training step size is set to 200, the initial learning rate is set to 0.02, and the batch_size is set to 1;

[0073] (2) Use the k-means clustering algorithm to obtain K target anchor values ​​and modify the original anchor box to the new anchor box after clustering;

[0074] (3) The input image is first subjected to Mosaic data enhancement and image normalization, and the image used for training is normalized to a size of 640×640;

[0075] (4) The Backbone part slices and convolves the image to obtain a feature map;

[0076] (5) The feature information of the feature map is extracted through the Neck part. In the Neck part, the image is first upsampled to extract image features, and then downsampled to fuse feature information of different scales. Finally, the detection result is output from the output end to obtain the detection model;

[0077] (6) Use the validation set to verify the performance of the model.

[0078] S200. Obtain two pictures containing body parts of the user taken by the user at two time periods, record them as the first image and the second image respectively, input the first image and the second image into the skin disease classification neural network model and the body part classification neural network model respectively, identify each body part and the lesion area of ​​each body part in the first image and the second image, and the type of skin disease corresponding to the lesion area, calculate the first area of ​​each lesion area and the second area of ​​the body part where the lesion area is located, calculate each first ratio of the first area to the second area of ​​each body part in the first image, and calculate each second ratio of the first area to the second area of ​​each body part in the second image.

[0079] It can be understood that the calculation of the first area and the second area is achieved by calculating the number of pixel points corresponding to the first area and the second area.

[0080] It should be noted here that the shooting time of the first image is earlier than the shooting time of the second image.

[0081] S300: Determine whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios.

[0082] Specifically, judging whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios includes:

[0083] Randomly select a skin disease suffered by the user, calculate the sum of the first ratios corresponding to the skin disease, and record it as the total proportion of the first lesion area of ​​the skin disease; and calculate the sum of the second ratios corresponding to the skin disease, and record it as the total proportion of the second lesion area of ​​the skin disease.

[0084] Determining whether the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease;

[0085] Determine whether the condition of the skin disease has improved based on the judgment result;

[0086] And calculate whether the condition of other types of skin diseases suffered by the user has improved according to the above method.

[0087] Specifically, judging whether the condition of the skin disease has improved according to the judgment result includes:

[0088] If the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease, it is determined that the skin disease has not improved, a prompt signal indicating that the skin disease has not improved is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​the skin disease are output;

[0089] If the total proportion of the second lesion area of ​​this type of skin disease is less than or equal to the total proportion of the first lesion area of ​​this type of skin disease, it is judged that this type of skin disease has improved, and a prompt signal regarding the improvement of this type of skin is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​this type of skin disease are output.

[0090] It can be understood that if the total proportion of the second lesion area of ​​this type of skin disease is greater than the total proportion of the first lesion area of ​​this type of skin disease, it proves that the total area of ​​this type of skin disease on the patient has increased, proving that this type of skin disease has not improved; otherwise, it proves that this type of skin disease has improved.

[0091] Specifically, the method further includes:

[0092] Whether the skin diseases suffered by the user have metastasized is determined according to the first ratios and the second ratios.

[0093] Specifically, judging whether the skin diseases suffered by the user have metastasized according to the first ratios and the second ratios includes:

[0094] Randomly select a skin disease suffered by the user, calculate the difference between the second ratio and the first ratio of each body part corresponding to the skin disease, respectively determine whether the difference is greater than a first preset threshold, and determine whether the skin disease suffered by the user has metastasized to each body part according to the determination result;

[0095] Calculate whether the user's other types of skin diseases have been transferred according to the above method.

[0096] Specifically, judging whether the skin disease of the user has metastasized to various parts of the body according to the judgment result includes:

[0097] If the difference of a certain body part is greater than a first preset threshold, it is determined that the skin disease has metastasized to the body part, a prompt signal indicating that the skin disease has metastasized to the body part is output, and the name of the body part is output;

[0098] If the difference of a certain body part is less than or equal to the first preset threshold, it is determined that the type of skin disease has not metastasized to the body part, a prompt signal is output indicating that the type of skin disease has not metastasized to the body part, and the name of the body part is output.

[0099] It can be understood that if the difference in a certain body part is greater than the first preset threshold, since the shooting time of the first image is earlier than the shooting time of the second image, it proves that the area of ​​the lesion region of this type of skin disease in this body part has increased, proving that this type of skin disease has metastasized to this body part, and a prompt signal about the metastasis of this type of skin disease to this body part is output, and the name of the body part is output.

[0100] It can be understood that if the difference in a certain body part is less than or equal to the first preset threshold, since the shooting time of the first image is earlier than the shooting time of the second image, it proves that the area of ​​the lesion region of this type of skin disease in this body part has not increased, and proves that this type of skin disease has not metastasized to this body part. A prompt signal about the metastasis of this type of skin disease in this body part is output, and the name of the body part is output.

[0101] It can be understood that the present invention can achieve synchronous intelligent judgment on whether the various skin diseases suffered by the user have improved and whether metastasis has occurred, only by obtaining two pictures containing the user's body parts taken by the user at two time periods. It greatly reduces the doctor's diagnostic workload and diagnostic accuracy, saves the doctor's diagnostic time to a great extent, improves the diagnostic efficiency, greatly improves the intelligence and usability of the present invention, and greatly expands the application scenarios of the present invention.

[0102] See also Figure 2 The present invention provides another embodiment, which provides a skin disease auxiliary diagnosis system based on deep learning, and the skin disease auxiliary diagnosis system based on deep learning includes:

[0103] The acquisition module 100 is used to acquire different skin lesion pictures; or to acquire different body part pictures; or to acquire two pictures of the user's body parts taken by the user at two periods;

[0104] The control module 200 is used to scale skin lesion images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a skin disease classification neural network model; or to scale body part images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a body part classification neural network model; or to record two images of the user's body parts taken by the user at two periods as a first image and a second image, respectively, and input the first image and the second image into a skin disease classification neural network model and a body part classification neural network model, respectively, identify each body part and the lesion area of ​​each body part in the first image and the second image, and the type of skin disease corresponding to the lesion area, calculate the first area of ​​each lesion area and the second area of ​​the body part where the lesion area is located, calculate each first ratio of the first area to the second area of ​​each body part in the first image, and calculate each second ratio of the first area to the second area of ​​each body part in the second image; or to judge whether the condition of each skin disease suffered by the user has improved according to each first ratio and each second ratio.

[0105] It should be noted here that the present invention can realize the synchronous intelligent judgment of whether the various skin diseases suffered by the user have improved and whether metastasis has occurred, only by obtaining two pictures containing the user's body parts taken by the user at two time periods. This greatly reduces the doctor's diagnostic workload and diagnostic accuracy, saves the doctor's diagnostic time to a great extent, improves the diagnostic efficiency, greatly improves the intelligence and usability of the present invention, and greatly expands the application scenarios of the present invention.

[0106] In a preferred embodiment, the present application further provides an electronic device, the electronic device comprising:

[0107] A memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method for auxiliary diagnosis of skin diseases based on deep learning is implemented. The computer device can be broadly a server, a terminal, or any other electronic device with necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, a memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide necessary computing, processing and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. may be stored in or on the non-volatile storage medium. The internal memory may provide an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be used to connect and communicate with external devices through a network. When the computer program is executed by the processor, the steps of the method of the present invention are performed.

[0108] The present invention may be implemented as a computer-readable storage medium having a computer program stored thereon, which causes the steps of the method of an embodiment of the present invention to be executed when executed by a processor. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be performed by one or more computer devices or processors, and one or more other method steps / operations may be performed by one or more other computer devices or processors. One or more computer devices or processors may perform a single method step / operation, or perform two or more method steps / operations.

[0109] It will be appreciated by those skilled in the art that the method steps of the present invention may be performed by instructing related hardware such as a computer device or a processor through a computer program, and the computer program may be stored in a non-temporary computer-readable storage medium, which causes the steps of the present invention to be performed when the computer program is executed. Depending on the circumstances, any reference to memory, storage, database, or other media herein may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0110] It can be understood that the present invention realizes that by acquiring the file to be processed, it is possible to judge whether the text to be processed contains synonyms of preset sensitive words. The user only needs to confirm a preset sensitive word, and the words in the text to be processed that are semantically similar to the preset sensitive word and the semantic text corresponding to the word can be automatically updated to the sensitive word database, thereby realizing timely self-update and self-optimization of the sensitive word database. There is no need for the user to set multiple keywords or search for synonyms by himself, which greatly reduces the user's workload, greatly improves the effect of sensitive period shielding, greatly improves the user experience, greatly improves the intelligence and usability of the present invention, and greatly expands the application scenarios of the present invention.

[0111] It can be understood that the present invention can achieve synchronous intelligent judgment on whether the various skin diseases suffered by the user have improved and whether metastasis has occurred, only by obtaining two pictures containing the user's body parts taken by the user at two time periods. It greatly reduces the doctor's diagnostic workload and diagnostic accuracy, saves the doctor's diagnostic time to a great extent, improves the diagnostic efficiency, greatly improves the intelligence and usability of the present invention, and greatly expands the application scenarios of the present invention.

[0112] The various technical features described above can be combined arbitrarily. Although all possible combinations of these technical features are not described, any combination of these technical features should be considered to be covered by this specification as long as there is no contradiction in such combination.

[0113] The specific implementation of the present invention described above does not constitute a limitation on the protection scope of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A skin disease diagnosis method based on deep learning, characterized in that: The method comprises: S100, obtaining different skin lesion images, scaling the skin lesion images of different sizes to a preset size; performing data amplification on the scaled images according to a preset method; and training a neural network model for skin disease classification using the amplified images; obtaining different body part images, scaling the body part images of different sizes to a preset size; performing data amplification on the scaled images according to a preset method; and training a neural network model for body part classification using the amplified images; S200, obtaining two pictures including body parts of the user taken by the user at two periods, respectively recorded as a first image and a second image, inputting the first image and the second image into a skin disease classification neural network model and a body part classification neural network model respectively, identifying each body part and the lesion area of ​​each body part in the first image and the second image, and the type of skin disease corresponding to the lesion area, calculating a first area of ​​each lesion area and a second area of ​​the body part where the lesion area is located, calculating each first ratio of the first area to the second area of ​​each body part in the first image, and calculating each second ratio of the first area to the second area of ​​each body part in the second image; S300: Determine whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios.

2. The skin disease auxiliary diagnosis method based on deep learning according to claim 1 is characterized in that: The body parts include: Face, chest, arms, legs, abdomen, back.

3. The skin disease auxiliary diagnosis method based on deep learning according to claim 1 is characterized in that: The step of judging whether the skin diseases suffered by the user have improved according to the first ratios and the second ratios includes: Randomly select a skin disease suffered by the user, calculate the sum of the first ratios corresponding to the skin disease, and record it as the total proportion of the first lesion area of ​​the skin disease; and calculate the sum of the second ratios corresponding to the skin disease, and record it as the total proportion of the second lesion area of ​​the skin disease; Determining whether the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease; Determine whether the condition of the skin disease has improved based on the judgment result; And calculate whether the condition of other types of skin diseases suffered by the user has improved according to the above method.

4. The skin disease auxiliary diagnosis method based on deep learning according to claim 3 is characterized in that: The step of judging whether the condition of the skin disease has improved according to the judgment result includes: If the total proportion of the second lesion area of ​​the skin disease is greater than the total proportion of the first lesion area of ​​the skin disease, it is determined that the skin disease has not improved, a prompt signal indicating that the skin disease has not improved is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​the skin disease are output; If the total proportion of the second lesion area of ​​this type of skin disease is less than or equal to the total proportion of the first lesion area of ​​this type of skin disease, it is judged that this type of skin disease has improved, and a prompt signal regarding the improvement of this type of skin is output, and the total proportion of the first lesion area and the total proportion of the second lesion area of ​​this type of skin disease are output.

5. The skin disease auxiliary diagnosis method based on deep learning according to claim 1, characterized in that: The step of performing data amplification on the scaled image according to a preset method includes: The scaled image is transformed by random cropping, random flipping, random rotation, and random deformation.

6. The skin disease auxiliary diagnosis method based on deep learning according to claim 1, characterized in that: The method further comprises: Whether the skin diseases suffered by the user have metastasized is determined according to the first ratios and the second ratios.

7. The skin disease auxiliary diagnosis method based on deep learning according to claim 6, characterized in that: The determining whether the skin diseases suffered by the user have metastasized according to the first ratios and the second ratios includes: Randomly select a skin disease suffered by the user, calculate the difference between the second ratio and the first ratio of each body part corresponding to the skin disease, respectively determine whether the difference is greater than a first preset threshold, and determine whether the skin disease suffered by the user has metastasized to each body part according to the determination result; Calculate whether the user's other types of skin diseases have been transferred according to the above method.

8. The skin disease auxiliary diagnosis method based on deep learning according to claim 7, characterized in that: The step of judging whether the skin disease of the user has metastasized to various parts of the body according to the judgment result includes: If the difference of a certain body part is greater than a first preset threshold, it is determined that the skin disease has metastasized to the body part, a prompt signal indicating that the skin disease has metastasized to the body part is output, and the name of the body part is output; If the difference of a certain body part is less than or equal to the first preset threshold, it is determined that the type of skin disease has not metastasized to the body part, a prompt signal is output indicating that the type of skin disease has not metastasized to the body part, and the name of the body part is output.

9. A skin disease auxiliary diagnosis system based on deep learning, characterized in that: include: An acquisition module, used to acquire different skin lesion images; or used to acquire different body part images; or used to obtain two pictures including body parts of the user taken by the user at two periods; A control module is used to scale skin lesion images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a skin disease classification neural network model; or to scale body part images of different sizes to a preset size; perform data amplification on the scaled images according to a preset method; and use the amplified images to train a body part classification neural network model; or to record two images of the user's body parts taken by the user at two periods as a first image and a second image, respectively, and input the first image and the second image into a skin disease classification neural network model and a body part classification neural network model, respectively, identify each body part and the lesion area of ​​each body part in the first image and the second image, and the skin disease type corresponding to the lesion area, calculate the first area of ​​each lesion area and the second area of ​​the body part where the lesion area is located, calculate each first ratio of the first area to the second area of ​​each body part in the first image, and calculate each second ratio of the first area to the second area of ​​each body part in the second image; or to judge whether the condition of each skin disease suffered by the user has improved according to each first ratio and each second ratio.

10. An electronic device, characterized in that: include: Memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the deep learning-based auxiliary diagnosis method for skin diseases according to any one of claims 1 to 8 is implemented.

Images