Skin lesion monitoring and diagnosing system
By designing a skin lesion monitoring and diagnosis system that integrates user interface, image processing and machine learning models, it solves the misdiagnosis, misdiagnosis and data security problems of traditional skin lesion diagnosis, and achieves efficient and accurate diagnosis and data protection.
Patent Information
- Application Number
- CN202410947863.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional skin lesions diagnosis relies on doctors’ experience, with high rates of misdiagnosis and missed diagnosis, low efficiency of manual image analysis, difficult to guarantee patient data security and privacy, and the existing system is complex and unfriendly.
A skin lesion monitoring and diagnosis system including user interface, image capture module, image preprocessing module, feature extraction module, machine learning model, diagnostic result generator, data storage system, medical database and feedback and update module was designed, using The deep learning model analyzes image features, generates preliminary diagnostic reports, and ensures data security through encryption technology.
It improves the accuracy and efficiency of diagnosis, increases patient accessibility, ensures data security, and lowers technical barriers, so that non-professionals can also use the system.
Smart Images

Figure CN118902390A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical health, in particular to the field of dermatology and primary health care, and more particularly to a skin lesion monitoring and diagnosis system. Background Art
[0002] Traditional dermatology diagnosis of skin lesions relies on the doctor's experience and intuition, which may lead to misdiagnosis or missed diagnosis. Professional dermatologists may be difficult to obtain in some areas, resulting in many patients not being able to receive a timely diagnosis. The existing technology uses a manual image analysis process that is time-consuming and inefficient, which is not conducive to rapid diagnosis and treatment. In addition, during the image transmission and storage process, the security and privacy of patient data are difficult to be fully guaranteed. Existing skin lesion monitoring systems may be too complex and unfriendly to non-professionals. Summary of the invention
[0003] The object of the present invention is to provide a skin lesion monitoring and diagnosis system.
[0004] To achieve the above object, the present invention is implemented according to the following technical solutions:
[0005] The present invention includes a user interface, an image capture module, an image preprocessing module, a feature extraction module, a machine learning model, a diagnosis result generator, a data storage system, a medical database and a feedback and update module. The control signal input end of the user interface is connected to the image capture module, the data output end of the image capture module is connected to the image preprocessing module, the data output end of the image preprocessing module is connected to the feature extraction module, the data output end of the feature extraction module is connected to the diagnosis result generator through the machine learning model, the data output end of the diagnosis result generator is connected to the data storage system, the data storage system is connected to the feedback and update module in combination with the medical database, and the feedback data output end of the feedback and update module is connected to the user interface.
[0006] The user interface is used to upload images and receive diagnosis results;
[0007] The image capture module allows a user to take images of skin lesions using a high-resolution camera;
[0008] The image preprocessing module processes the captured images, including denoising, adjusting brightness and contrast, cropping, etc., to improve image quality;
[0009] The feature extraction module extracts features that are helpful for lesion identification from the preprocessed images;
[0010] The machine learning model analyzes features using a trained deep learning model and compares them with known lesion patterns.
[0011] The diagnosis result generator generates a preliminary diagnosis report based on the analysis results of the machine learning model.
[0012] The data storage system securely stores user data and diagnostic results.
[0013] Medical databases contain databases of known lesion patterns and related medical information, which are used to train and validate machine learning models.
[0014] The feedback and update module collects user feedback to improve system performance and regularly updates machine learning models.
[0015] The image preprocessing module is specifically:
[0016] Denoising: Use filters to remove noise from an image:
[0017] Gaussian filter:
[0018]
[0019] Contrast enhancement: Enhance the contrast of the image to make the lesion features more obvious;
[0020] Contrast limited adaptive histogram equalization;
[0021] Normalize: scale the image to a uniform size;
[0022] Rotation correction: If the image is tilted, it is corrected through feature point detection and affine transformation;
[0023] Segmentation: Separate the lesion area from the background;
[0024] Threshold segmentation:
[0025] The feature extraction is specifically as follows:
[0026] Color features: Use color histograms or color moments to describe the color distribution of lesions;
[0027] Texture features: Use gray-level co-occurrence matrix to describe the texture of the lesion;
[0028] Shape features: Calculate the boundaries of the lesion and extract shape descriptors such as circularity and ellipticity;
[0029] Edge features: Canny edge detector is used to extract edge information of lesions.
[0030] The machine learning model is specifically:
[0031] Data preprocessing: normalize the labeled image data;
[0032] Model architecture: Design CNN architecture, including convolutional layer, pooling layer, and fully connected layer;
[0033] Loss function: Select the cross entropy loss function;
[0034] Optimizer: Select the Adam optimizer optimization algorithm;
[0035] Training: Use the back-propagation algorithm to train the model;
[0036] Validation and testing: Use the validation set to evaluate the model performance for testing.
[0037] The machine learning model generates a preliminary diagnosis result based on the output of the machine learning model:
[0038] Threshold determination: Set a threshold for the model output to determine the lesion type;
[0039] Result mapping: Mapping model output to specific lesion types;
[0040] Uncertainty estimation: Provide an uncertainty estimate for the diagnosis result;
[0041] Report Generation: Generates a report containing diagnostic results and uncertainty estimates.
[0042] The beneficial effects of the present invention are:
[0043] The present invention is a skin lesion monitoring and diagnosis system. Compared with the prior art, the present invention has the following technical effects:
[0044] Improve accuracy: Analyze lesion images through machine learning models to improve the accuracy of diagnosis.
[0045] Increased accessibility: The system can be accessed via mobile devices or the Internet, allowing patients in remote areas to receive professional skin lesion analysis.
[0046] Improve efficiency: The automated image analysis process significantly shortens diagnosis time and improves medical efficiency.
[0047] Ensure data security: Use encryption technology and security protocols to ensure the security and privacy of user data.
[0048] Lower technical threshold: The user-friendly interface design makes it easy for non-professionals to use the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 It is a system structure principle block diagram of the present invention. DETAILED DESCRIPTION
[0050] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.
[0051] like Figure 1 As shown: the present invention includes a user interface, an image capture module, an image preprocessing module, a feature extraction module, a machine learning model, a diagnosis result generator, a data storage system, a medical database and a feedback and update module, the control signal input end of the user interface is connected to the image capture module, the data output end of the image capture module is connected to the image preprocessing module, the data output end of the image preprocessing module is connected to the feature extraction module, the data output end of the feature extraction module is connected to the diagnosis result generator through the machine learning model, the data output end of the diagnosis result generator is connected to the data storage system, the data storage system is connected to the feedback and update module in combination with the medical database, and the feedback data output end of the feedback and update module is connected to the user interface.
[0052] The user interface is used to upload images and receive diagnosis results;
[0053] The image capture module allows a user to take images of skin lesions using a high-resolution camera;
[0054] The image preprocessing module processes the captured images, including denoising, adjusting brightness and contrast, cropping, etc., to improve image quality;
[0055] The feature extraction module extracts features that are helpful for lesion identification from the preprocessed images;
[0056] The machine learning model analyzes features using a trained deep learning model and compares them with known lesion patterns.
[0057] The diagnosis result generator generates a preliminary diagnosis report based on the analysis results of the machine learning model.
[0058] The data storage system securely stores user data and diagnostic results.
[0059] Medical databases contain databases of known lesion patterns and related medical information, which are used to train and validate machine learning models.
[0060] The feedback and update module collects user feedback to improve system performance and regularly updates machine learning models.
[0061] The image preprocessing module is specifically:
[0062] Denoising: Use filters to remove noise from an image:
[0063] Gaussian filter:
[0064]
[0065] Contrast enhancement: Enhance the contrast of the image to make the lesion features more obvious;
[0066] Contrast limited adaptive histogram equalization;
[0067] Normalize: scale the image to a uniform size;
[0068] Rotation correction: If the image is tilted, it is corrected through feature point detection and affine transformation;
[0069] Segmentation: Separate the lesion area from the background;
[0070] Threshold segmentation: The feature extraction is specifically as follows:
[0071] Color features: Use color histograms or color moments to describe the color distribution of lesions;
[0072] Texture features: Use gray-level co-occurrence matrix to describe the texture of the lesion;
[0073] Shape features: calculate the boundary of the lesion and extract shape descriptors such as circularity and ellipticity; Edge features: use the Canny edge detector to extract the edge information of the lesion.
[0074] The machine learning model is specifically:
[0075] Data preprocessing: normalize the labeled image data;
[0076] Model architecture: Design CNN architecture, including convolutional layer, pooling layer, and fully connected layer;
[0077] Loss function: Select the cross entropy loss function;
[0078] Optimizer: Select the Adam optimizer optimization algorithm;
[0079] Training: Use the back-propagation algorithm to train the model;
[0080] Validation and testing: Use the validation set to evaluate the model performance for testing.
[0081] The machine learning model generates a preliminary diagnosis result based on the output of the machine learning model:
[0082] Threshold determination: Set a threshold for the model output to determine the lesion type;
[0083] Result mapping: Mapping model output to specific lesion types;
[0084] Uncertainty estimation: Provide an uncertainty estimate for the diagnosis result;
[0085] Report Generation: Generates a report containing diagnostic results and uncertainty estimates.
[0086] Example: Development of a skin lesion monitoring and diagnosis system
[0087] 1. Demand Analysis
[0088] Target users: Primary care physicians and patients.
[0089] Functional requirements: Users upload images of skin lesions, and the system provides a preliminary diagnosis of the lesion type and severity.
[0090] 2. Hardware Selection
[0091] Camera: Choose a smartphone camera with a resolution of 12MP.
[0092] Light source: Use a ring LED light to ensure even lighting.
[0093] 3. System development
[0094] User Interface: Design a simple web interface where users can upload images and receive feedback.
[0095] Image Capture: Develop a mobile application that allows the user to capture images of lesions using the phone camera.
[0096] 4. Machine Learning Model Development
[0097] Data collection: Collect 1000 annotated skin lesion images.
[0098] Model training: Use these images to train a 5-layer CNN model, including 3 convolutional layers and 2 fully connected layers.
[0099] 5. System Integration
[0100] Module integration: Integrate image capture, preprocessing, feature extraction, model analysis, and result generation into one system.
[0101] 6. Data Security
[0102] Encryption: Use SSL / TLS to encrypt user data transmission.
[0103] 7. System Testing
[0104] Unit testing: Test each module.
[0105] User testing: Invite 20 doctors and 50 patients to conduct testing and collect feedback.
[0106] Detailed implementation steps:
[0107] Image preprocessing:
[0108] Denoising: Use a Gaussian filter to denoise the image.
[0109] Contrast Enhancement: Apply the CLAHE algorithm to enhance the image contrast.
[0110] Normalize: Scale the image to a uniform size, such as 256x256 pixels.
[0111] Feature extraction:
[0112] Color features: Calculate the color histogram of the image.
[0113] Texture features: Texture features are calculated using GLCM.
[0114] Shape features: Use edge detection algorithms to extract lesion shape features.
[0115] Machine Learning Models:
[0116] Data preprocessing: Normalize image data to the range [0,1].
[0117] Model Architecture:
[0118] Input layer: accepts 256x256x3 image data.
[0119] Convolutional layer 1: 32 3x3 convolution kernels, ReLU activation.
[0120] Pooling layer 1: 2x2 max pooling.
[0121] Convolutional layer 2: 64 3x3 convolution kernels, ReLU activation.
[0122] Pooling layer 2: 2x2 max pooling.
[0123] Fully connected layer 1: 512 neurons, ReLU activation.
[0124] Output layer: Use the softmax function to output the probability of the lesion type.
[0125] Loss function: Use the cross entropy loss function.
[0126] Optimizer: Use Adam optimizer for model training.
[0127] Diagnostic Result Generator:
[0128] Threshold determination: Set the threshold of the model output probability to 0.5 to determine the lesion type. Result mapping: Map the lesion to the most likely type based on the probability of the model output. Uncertainty estimation: Calculate the probability distribution of the model output and provide uncertainty estimation. Implementation results:
[0129] User feedback: According to feedback from user testing, the system accuracy is about 85%.
[0130] Market Acceptability: The system received positive reviews among primary care physicians, who considered it a useful adjunct tool.
[0131] This example provides a high-level view of how to transform a technical solution into a specific project. In actual operation, each step requires detailed planning, execution, and testing to ensure the reliability and effectiveness of the system.
[0132] The technical solution of the present invention is not limited to the above-mentioned specific embodiments. All technical variations made according to the technical solution of the present invention fall within the protection scope of the present invention.
Claims
1. A skin lesion monitoring and diagnosis system, characterized in that: It includes a user interface, an image capture module, an image preprocessing module, a feature extraction module, a machine learning model, a diagnosis result generator, a data storage system, a medical database and a feedback and update module. The control signal input end of the user interface is connected to the image capture module, the data output end of the image capture module is connected to the image preprocessing module, the data output end of the image preprocessing module is connected to the feature extraction module, the data output end of the feature extraction module is connected to the diagnosis result generator through the machine learning model, the data output end of the diagnosis result generator is connected to the data storage system, the data storage system is connected to the feedback and update module in combination with the medical database, and the feedback data output end of the feedback and update module is connected to the user interface.
2. The skin lesion monitoring and diagnosis system according to claim 1, characterized in that: The user interface is used to upload images and receive diagnosis results; The image capture module allows a user to take images of skin lesions using a high-resolution camera; The image preprocessing module processes the captured images, including denoising, adjusting brightness and contrast, cropping, etc., to improve image quality; The feature extraction module extracts features that are helpful for lesion identification from the preprocessed images; The machine learning model analyzes features using a trained deep learning model and compares them with known lesion patterns. The diagnosis result generator generates a preliminary diagnosis report based on the analysis results of the machine learning model. The data storage system securely stores user data and diagnostic results. Medical databases contain databases of known lesion patterns and related medical information, which are used to train and validate machine learning models. The feedback and update module collects user feedback to improve system performance and regularly updates machine learning models.
3. The skin lesion monitoring and diagnosis system according to claim 2, characterized in that: The image preprocessing module is specifically: Denoising: Use filters to remove noise from an image: Gaussian filter: Contrast enhancement: Enhance the contrast of the image to make the lesion features more obvious; Contrast limited adaptive histogram equalization; Normalize: scale the image to a uniform size; Rotation correction: If the image is tilted, it is corrected through feature point detection and affine transformation; Segmentation: Separate the lesion area from the background; Threshold segmentation:
4. The skin lesion monitoring and diagnosis system according to claim 2, characterized in that: The feature extraction is specifically as follows: Color features: Use color histograms or color moments to describe the color distribution of lesions; Texture features: Use gray-level co-occurrence matrix to describe the texture of the lesion; Shape features: Calculate the boundaries of the lesion and extract shape descriptors such as circularity and ellipticity; Edge features: Canny edge detector is used to extract edge information of lesions.
5. The skin lesion monitoring and diagnosis system according to claim 2, characterized in that: The machine learning model is specifically: Data preprocessing: normalize the labeled image data; Model architecture: Design CNN architecture, including convolutional layer, pooling layer, and fully connected layer; Loss function: Select the cross entropy loss function; Optimizer: Select the Adam optimizer optimization algorithm; Training: Use the back-propagation algorithm to train the model; Validation and testing: Use the validation set to evaluate the model performance for testing.
6. The skin lesion monitoring and diagnosis system according to claim 2, characterized in that: The machine learning model generates a preliminary diagnosis result based on the output of the machine learning model: Threshold determination: Set a threshold for the model output to determine the lesion type; Result mapping: Mapping model output to specific lesion types; Uncertainty estimation: Provide an uncertainty estimate for the diagnosis result; Report Generation: Generates a report containing diagnostic results and uncertainty estimates.
Citation Information
Cited By
Knee joint image intelligent processing method based on MRI and application of knee joint image intelligent processing method in meniscus posterior root tearing diagnosis system
CN120107184A
Skin disease auxiliary diagnosis equipment based on computer and system thereof
CN120477724A
