Intelligent paronychia detection system

Abstract

The invention discloses an intelligent paronychia detection system, which comprises a data acquisition module, and is characterized in that the data acquisition module is used for collecting a large number of toe photos of paronychia patients as training data; the preprocessing module is used for preprocessing the collected photos; the model construction module is used for constructing a paronychia detection model based on a deep learning technology; the training module is used for training the model by using the pre-processed photos, and improving the paronychia feature recognition capability of the model by continuously optimizing model parameters, and finally obtaining an optimal paronychia detection model; the paronychia detection method has the beneficial effects that by introducing the deep learning technology, automation and intelligence of paronychia detection are realized, and the detection efficiency and accuracy are remarkably improved; through learning and training of a large number of samples, the model can accurately identify various features of paronychia, including subtle lesion signs, and the detection precision is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of paronychia detection, and in particular relates to an intelligent paronychia detection system. Background Art

[0002] Paronychia is a common foot disease that mainly occurs around the toenails and is caused by bacterial infection. It manifests as redness, swelling, and pain proximal to the nail groove on one or both sides of the toenail. Pus spots and pus discharge may follow. In severe cases, it can lead to hyperplasia of granulation tissue and even cause the entire nail to float and fall off.

[0003] Testing can detect signs of paronychia early to avoid further progression of the disease. Depending on the severity of the disease, doctors can develop an appropriate treatment plan. If paronychia is not treated promptly, the infection may spread to the nail bed or surrounding tissues, and even cause systemic infection. Testing can detect and control the infection in a timely manner to prevent complications. Testing helps doctors provide patients with personalized care advice. During treatment, regular testing can evaluate treatment effectiveness and adjust treatment plans in a timely manner. Paronychia is prone to recurrence, and testing can help understand the causes and risk factors for recurrence so that appropriate preventive measures can be taken.

[0004] Traditional paronychia detection methods mainly rely on doctors' direct observation and empirical judgment. This method is easily affected by factors such as the external environment and differences in doctors' experience, making it difficult to ensure the accuracy and consistency of the test results. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent paronychia detection system, which realizes automated and high-precision paronychia detection by pre-shooting and processing paronychia photos, constructing and training a paronychia detection model.

[0006] To achieve the above object, the present invention provides the following technical solution: a paronychia intelligent detection system, comprising

[0007] A data collection module, the data collection module is used to collect a large number of toe photos of patients with paronychia as training data;

[0008] A preprocessing module, wherein the preprocessing module preprocesses the collected photos;

[0009] A model building module, wherein the model building module builds a paronychia detection model based on deep learning technology;

[0010] A training module, wherein the training module uses the preprocessed photos to train the model, and improves the model's ability to recognize paronychia features by continuously optimizing the model parameters, and ultimately obtains an optimal paronychia detection model;

[0011] A real-time detection module, wherein when the real-time detection module is used for detection, a high-definition camera is used to take photos of the patient's toes to be detected, and a pre-processing operation is performed on the taken photos;

[0012] An import module, wherein the import module is used to input the processed photo into the optimal paronychia detection model, and the optimal paronychia detection model automatically analyzes the image features and gives a judgment result on whether the patient suffers from paronychia and the severity of the disease;

[0013] The result display and feedback module displays the test results to doctors or patients in a graphical interface, provides user interaction functions, allows doctors to confirm or correct the test results, and collects user feedback to further optimize the model.

[0014] As a preferred technical solution of the present invention, a large number of toe photos of patients with paronychia are collected as training data, including photos of normal toes and photos of paronychia of different degrees.

[0015] As a preferred technical solution of the present invention, the collected photos are preprocessed, including denoising, contrast enhancement, and edge detection.

[0016] As a preferred technical solution of the present invention, the denoising includes a statistical filtering method and a wavelet transform method; the contrast enhancement includes a histogram equalization and contrast stretching; and the edge detection includes a gradient-based method and a machine learning-based method.

[0017] As a preferred technical solution of the present invention, the test results are displayed to doctors or patients in a graphical interface, including highlighting of the lesion area and a condition assessment report.

[0018] As a preferred technical solution of the present invention, the method for constructing a paronychia detection model is as follows:

[0019] Collect a large number of foot images of patients with paronychia and healthy people;

[0020] Annotate the collected images to clearly indicate which images show symptoms of paronychia and which images are healthy;

[0021] Remove blurry, duplicate or poor quality images; increase the amount of data through rotation, scaling, cropping, and flipping operations to improve the generalization ability of the model; adjust the pixel value range of the image to make it suitable for the input requirements of the model;

[0022] Select convolutional neural network for model building.

[0023] As a preferred technical solution of the present invention, the method for optimizing model parameters includes parameter adjustment strategy, gradient descent method and its variants.

[0024] As a preferred technical solution of the present invention, the variants include batch gradient descent, stochastic gradient descent, and mini-batch gradient descent.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] By introducing deep learning technology, the automation and intelligence of paronychia detection has been realized, which significantly improves the detection efficiency and accuracy;

[0027] By learning and training from a large number of samples, the model can accurately identify various characteristics of paronychia, including subtle signs of lesions, thus improving the accuracy of detection;

[0028] Compared with the traditional manual detection method, the present invention reduces the influence of external factors such as light and angle on the detection results through image processing technology;

[0029] It adopts a combination of multiple modules, each with clear functions and independent of each other, which is easy to maintain and upgrade. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a system principle diagram of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Example 1

[0033] See also Figure 1 , which is the first embodiment of the present invention, and provides an intelligent paronychia detection system, comprising

[0034] Data collection module, the data collection module is used to collect a large number of toe photos of patients with paronychia as training data, including normal toe photos and photos of paronychia of different degrees;

[0035] Preprocessing module: The preprocessing module preprocesses the collected photos, including denoising, contrast enhancement, edge detection, and improves image quality to facilitate subsequent model training. Denoising refers to removing noise from the image to make the image clearer. Commonly used denoising methods include:

[0036] Statistical filtering method:

[0037] Mean filtering: Use the average gray value of pixels in the image to eliminate noise, which has a good effect on Gaussian noise;

[0038] Median filtering: Use the median value of the pixel in the area window around the pixel to eliminate noise;

[0039] Adaptive filtering methods: such as adaptive weighted median filtering and bilateral filtering, which can adjust filter parameters according to local characteristics of the image to better preserve image details;

[0040] Wavelet transform method:

[0041] Wavelet transform decomposes the image into frequency bands of different scales and uses the correlation between frequency bands to eliminate noise;

[0042] Contrast enhancement refers to making the difference between the brightness values ​​of different areas in an image more obvious through a series of image processing operations to improve the visual effect and readability of the image; commonly used methods include:

[0043] Histogram Equalization:

[0044] By adjusting the gray value distribution of the image, the pixel points of each gray level in the image are distributed more evenly, thereby enhancing the contrast of the image;

[0045] Adaptive Histogram Equalization:

[0046] Divide the image into many small areas and perform histogram equalization on each area to improve the enhancement effect while retaining the image's detail information;

[0047] Contrast stretching:

[0048] By linearly stretching the pixel values ​​of the image, the range of pixel values ​​in the original image is mapped to a new range, thereby enhancing the contrast of the image; linear stretching is an improved method, and the parameters can be flexibly adjusted;

[0049] Local contrast enhancement:

[0050] Apply the contrast enhancement method to the local area, and adaptively enhance the contrast of each pixel by analyzing the local features of the image (such as edges, textures, etc.);

[0051] Nonlinear enhancement method:

[0052] Such as logarithmic transformation, gamma transformation, histogram matching, etc., which can enhance different gray levels of the image to different degrees according to actual needs;

[0053] Edge model based methods:

[0054] Hough transform: represents the edges in the image as curves or straight lines in parameter space and detects edges by finding intersection points;

[0055] Edge tracking algorithm: an edge detection method based on pixel domain relationships, which forms a complete edge by continuously tracking edge points;

[0056] Machine learning based methods:

[0057] Use deep learning models (such as convolutional neural networks (CNNs)) for edge detection, and automatically learn image features and detect edges by training the model;

[0058] Model building module, the model building module builds a paronychia detection model based on deep learning technology;

[0059] The training module uses the pre-processed photos to train the model, and continuously optimizes the model parameters to improve the model's ability to recognize paronychia features, ultimately obtaining the optimal paronychia detection model;

[0060] Real-time detection module: When the real-time detection module is used for detection, a high-definition camera is used to take photos of the patient's toes to be detected, and the taken photos are pre-processed;

[0061] An import module is used to input the processed photos into the optimal paronychia detection model. The optimal paronychia detection model automatically analyzes the image features and gives a judgment result on whether the patient has paronychia and the severity of the disease.

[0062] The result display and feedback module displays the test results to doctors or patients in a graphical interface, including highlighting of lesion areas and condition assessment reports. It provides user interaction functions, allows doctors to confirm or correct the test results, and collects user feedback to further optimize the model.

[0063] A large number of toe photos of patients with paronychia were collected as training data, including photos of normal toes and photos of paronychia of different degrees.

[0064] Example 2

[0065] See also Figure 1 , which is the second embodiment of the present invention, is based on the previous embodiment, except that:

[0066] The method for constructing a paronychia detection model is as follows:

[0067] Collect a large number of foot images (especially toenail area) of patients with paronychia and healthy people; these images can come from hospitals, clinics, online medical platforms or be taken by patients themselves;

[0068] Label the collected images to clearly mark which images show symptoms of paronychia (such as redness, swelling, pain, pus, etc.) and which images are healthy, which usually requires the participation of medical experts;

[0069] Remove blurry, duplicate or poor quality images; increase the amount of data by rotating, scaling, cropping, flipping and other operations to improve the generalization ability of the model; adjust the pixel value range of the image to make it suitable for the input requirements of the model; if the image contains multiple toes or irrelevant areas, it needs to be cropped to only contain the toenails and their surrounding areas;

[0070] For image classification tasks, commonly used deep learning architectures include convolutional neural networks (CNNs) such as VGG, ResNet, Inception, etc. These networks have performed well in the field of image recognition;

[0071] Considering that training a deep learning model from scratch requires a lot of data and computing resources, one can consider using a pre-trained model (such as a model trained on ImageNet) for transfer learning. By fine-tuning the last few layers of the pre-trained model, it can be adapted to the task of paronychia detection.

[0072] Methods for optimizing model parameters include

[0073] Gradient descent is one of the most commonly used optimization algorithms. It calculates the gradient of the loss function with respect to the model parameters and updates the parameters in the opposite direction of the gradient to minimize the loss function. Its variants include:

[0074] Batch gradient descent: Use the entire training set to calculate the gradient, which is stable but computationally intensive.

[0075] Stochastic gradient descent: Only one sample is used to calculate the gradient at a time, and the update speed is fast;

[0076] Mini-batch gradient descent: A compromise between the above two methods, using a small batch of samples to calculate the gradient each time, which ensures both the update speed and relative stability;

[0077] Adaptive learning rate algorithm

[0078] In order to deal with the problems that may be caused by fixed learning rate in gradient descent method (such as slow convergence due to too small learning rate, and oscillation or even divergence due to too large learning rate), a variety of adaptive learning rate algorithms have emerged, such as:

[0079] AdaGrad: Adaptively adjusts the learning rate of each parameter based on the historical sum of squared gradients, giving smaller learning rates to frequently occurring parameters and larger learning rates to infrequently occurring parameters;

[0080] RMSprop: Similar to AdaGrad, it introduces a decay factor to limit the impact of the sum of squared historical gradients, thus avoiding the problem of premature decay of the learning rate.

[0081] Parameter Tuning Strategy

[0082] Grid search: Through exhaustive search, all possible parameter combinations are tried within the specified parameter range to find the optimal parameter combination;

[0083] Random search: Unlike grid search, random search randomly selects parameter combinations within the specified parameter range to try, and finds a better parameter combination in a smaller number of attempts;

[0084] Bayesian optimization: Use Bayes’ theorem to estimate the posterior distribution of the objective function and select the next evaluation point based on the posterior distribution. This method is particularly useful when the parameter space is large or the evaluation cost is high.

[0085] Regularization

[0086] In order to prevent the model from overfitting, a regularization term is added to the loss function to constrain the complexity of the model parameters.

[0087] Although the embodiments of the present invention have been shown and described, as detailed above, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Paronychia intelligent detection system, Features: include A data collection module, the data collection module is used to collect a large number of toe photos of patients with paronychia as training data; A preprocessing module, wherein the preprocessing module preprocesses the collected photos; A model building module, wherein the model building module builds a paronychia detection model based on deep learning technology; A training module, wherein the training module uses the preprocessed photos to train the model, and improves the model's ability to recognize paronychia features by continuously optimizing the model parameters, and ultimately obtains an optimal paronychia detection model; A real-time detection module, wherein when the real-time detection module is used for detection, a high-definition camera is used to take photos of the patient's toes to be detected, and a pre-processing operation is performed on the taken photos; An import module, wherein the import module is used to input the processed photo into the optimal paronychia detection model, and the optimal paronychia detection model automatically analyzes the image features and gives a judgment result on whether the patient suffers from paronychia and the severity of the disease; The result display and feedback module displays the test results to doctors or patients in a graphical interface, provides user interaction functions, allows doctors to confirm or correct the test results, and collects user feedback to further optimize the model.

2. The intelligent paronychia detection system according to claim 1, Features: A large number of toe photos of patients with paronychia were collected as training data, including photos of normal toes and photos of paronychia of different degrees.

3. The intelligent paronychia detection system according to claim 1, Features: The collected photos are preprocessed, including denoising, contrast enhancement, and edge detection.

4. The intelligent paronychia detection system according to claim 3, Features: The denoising includes statistical filtering and wavelet transform; the contrast enhancement includes histogram equalization and contrast stretching; the edge detection includes a gradient-based method and a machine learning-based method.

5. The intelligent paronychia detection system according to claim 1, Features: The test results are displayed to doctors or patients in a graphical interface, including highlighted lesion areas and condition assessment reports.

6. The intelligent paronychia detection system according to claim 1, Features: The method for constructing a paronychia detection model is as follows: Collect a large number of foot images of patients with paronychia and healthy people; Annotate the collected images to clearly indicate which images show symptoms of paronychia and which images are healthy; Remove blurry, duplicate or poor quality images; increase the amount of data through rotation, scaling, cropping, and flipping operations to improve the generalization ability of the model; adjust the pixel value range of the image to make it suitable for the input requirements of the model; Select convolutional neural network for model building.

7. The intelligent paronychia detection system according to claim 1, Features: Methods for optimizing model parameters include parameter tuning strategies, gradient descent, and its variants.

8. The intelligent paronychia detection system according to claim 7, Features: The variants include batch gradient descent, stochastic gradient descent, and mini-batch gradient descent.

Images