Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Eye fundus photo classification method and eye fundus image processing method and system

A classification method and photo technology, applied in the field of medical image classification and detection, can solve the problems of low accuracy and low efficiency, and achieve the effect of high effect, solving overfitting and occupying less computer resources.

Pending Publication Date: 2022-07-01
BEIHANG UNIV
View PDF0 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] A method and system for classifying fundus photos proposed by the present invention can solve the technical problems of low efficiency and low accuracy of the classification method based on fundus photos in the existing detection of chronic kidney disease

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Eye fundus photo classification method and eye fundus image processing method and system
  • Eye fundus photo classification method and eye fundus image processing method and system
  • Eye fundus photo classification method and eye fundus image processing method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0089] The fundus photo classification method described in the embodiment of the present invention mainly uses image morphology and deep learning methods, uses the morphological method to preprocess the fundus photos, and then introduces a more suitable The focal loss loss function for the incidence problem and the training strategy for the class imbalance problem, an automatic classification method based on the DenseNet-101 network model is designed.

[0090] To achieve the above purpose of the invention, an embodiment of the present invention proposes a method for image preprocessing and automatic classification of fundus photos based on a deep learning algorithm, including the following steps:

[0091] (1) Through the preprocessing of the fundus photo, a standardized fundus photo is generated, such as figure 2 shown

[0092] 1.15) For each fundus photo, first extract the region of interest (ROI), extract the R channel component of the fundus photo, and binarize it;

[00...

Embodiment 2

[0143] In the prior art, a convolutional neural network model using an attention branch module is used to process fundus image data, but the attention branch module still has many deficiencies. First of all, the attention branch module refers to the practice of CAM, and replaces the fully connected layer connected with softmax to GAP. Compared with the fully connected layer that can be trained with parameters, GAP is only a pooling operation and cannot be adjusted, and the previous layer is added. The training pressure of the convolutional layer leads to slower global convergence. Secondly, the attention branch module receives the feature map output from the previous convolution layer, and compresses the feature map information into CAM through ordinary convolution and 1x1x1 convolution; on the other branch, the attention branch module uses 1x1 convolution and The GAP method constrains the loss function. In fact, the two components of 1x1 convolution and GAP can directly expor...

Embodiment approach

[0145] Specifically, see Figure 5 , according to a preferred embodiment of the present invention, a fundus image processing method is provided, which is characterized by comprising the following steps:

[0146] Constructing the first data set, including the preprocessing of the original fundus image and the data screening of the preprocessed fundus image, the preprocessing of the original fundus image includes data desensitization, ROI extraction and contrast enhancement, and the preprocessed fundus image. The image data screening includes using the MobileNet v3 large model and the soft voting method to screen the preprocessed fundus images, clearing the fundus images with substandard quality, and constructing the first data set;

[0147]Building a convolutional neural network model with a CCAM module, the convolutional neural network model comprising a backbone network and the CCAM module, the backbone network comprising an input, a first convolutional layer, an MLP and a fi...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

According to the fundus photo classification method and the fundus image processing method and system, the technical problems that an existing fundus photo classification method is low in efficiency and large in error can be solved. Comprising the steps of obtaining and preprocessing a fundus photo to generate a standardized fundus picture; dividing the processed fundus photos into a training set and a test set, and performing data amplification on the training set; performing data preprocessing on the training set and the test set, and ensuring that each training data set (batch) input into the model in the training stage is uniform in category; integrating learning strategies, and training a branch neural network model; and carrying out model fusion to obtain a final detection model so as to realize classification of the fundus photos. The device is high in effect and excellent in automation degree; according to the method, features such as the optic cup / optic disk ratio and the arteriovenous ratio of the fundus photo are extracted, classification is carried out in combination with a machine learning method, the calculation speed is high, and few computer resources are occupied during operation.

Description

technical field [0001] The invention relates to the technical field of medical image classification and detection, in particular to a fundus photo classification method, a fundus image processing method and a system. Background technique [0002] Chronic kidney disease (CKD) is a chronic disease that affects global public health. The prevalence and incidence of chronic kidney disease is increasing year by year, and it is characterized by high morbidity, difficult treatment, high mortality and low awareness. Chronic kidney disease is often accompanied by cardiovascular diseases such as hypertension and diabetes. When chronic kidney disease gradually deteriorates, patients need kidney transplantation or long-term dependence on dialysis to maintain life. [0003] Studies have shown that early diagnosis and treatment of chronic kidney disease patients can effectively prevent further deterioration of the disease, and one of the means of prevention and treatment is to observe the...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06V10/764G06V10/50G06V10/80G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/241G06F18/25
Inventor 张冀聪王雄
Owner BEIHANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com