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

Anterior chamber angle image grading method fused with weak supervision metric learning

A technology of metric learning and grading method, applied in the field of medical image processing, can solve the problems of low accuracy and low efficiency, and achieve the effect of strengthening learning, reducing interference, and efficient diagnosis and treatment plan

Pending Publication Date: 2022-01-21
GUANGDONG POLYTECHNIC NORMAL UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the defects of low efficiency and low accuracy of the above-mentioned existing anterior chamber angle classification methods, the present invention provides an anterior chamber angle image classification method that integrates weakly supervised metric learning

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
  • Anterior chamber angle image grading method fused with weak supervision metric learning
  • Anterior chamber angle image grading method fused with weak supervision metric learning
  • Anterior chamber angle image grading method fused with weak supervision metric learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0057] Such as figure 1 As shown, a method for grading anterior chamber angle images that incorporates weakly supervised metric learning includes the following steps:

[0058] S1: Obtain anterior chamber angle images and perform image screening;

[0059] It should be noted that, at first, the images of the anterior chamber angle are collected, for example, at least 1000 images of the anterior chamber angle can be collected (such as figure 2 shown as the anterior chamber angle image), the collected images of the anterior chamber angle are screened, including: eliminating the images that cannot distinguish the grade of the anterior chamber angle due to objective conditions, and the images that cannot distinguish the grade of the anterior chamber angle due to objective conditions include : No images of anterior chamber angle structure, images with too strong light or too dark light with unclear structure of anterior chamber angle; images whose grade features are not obvious and...

Embodiment 2

[0076] In this embodiment, step S5 is described in detail based on the above steps.

[0077] S5: Use the training set and verification set to train and verify the constructed model to obtain the optimal network model;

[0078] Such as Figure 4 As shown, the specific process is:

[0079] S501: Select images for training from the training set, and perform data enhancement processing on the selected images;

[0080] The process of data enhancement is: perform random horizontal mirroring on the selected anterior chamber angle image with a probability of 0.5 (for example, randomly select 50% of the images to be enhanced for horizontal mirroring), and add salt and pepper noise to the image with a probability of 0.5 , where the signal-to-noise ratio is 0.95; the image is scaled; finally, the image is normalized and regularized, so that the image before the input model is in tensor format, and the value distribution is between [-1,1] ;

[0081] S502: Input the enhanced image data...

Embodiment 3

[0104] In this embodiment, the existing technology is divided into a traditional network and a weakly supervised network. The traditional network includes: VGG, GoogLeNet and ResNet. The traditional network consists of a large number of convolutional layers, pooling layers, batch normalization layers, and activation layers. Several classic models of convolutional neural networks; weakly supervised network methods are mainly divided into three types: consistency regularization, pseudo-label labeling, combined use of consistency regularization and pseudo-label labeling, these three methods are respectively provided by CCT, UPS , FixMatch as a representative method for comparative experiments, wherein Ours represents the method of the present invention, and the experimental results are shown in Table 1 and Table 2.

[0105] Table 1 Comparison table of classification performance on anterior chamber angle images and fundus images

[0106]

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

The invention discloses an anterior chamber angle image grading method fused with weak supervision metric learning. The anterior chamber angle image grading method comprises the steps of S1, obtaining anterior chamber angle images and carrying out image screening; S2, preprocessing the screened image to obtain an anterior chamber angle data set, wherein the anterior chamber angle data set comprises a training set, a verification set and a test set; S3, respectively constructing an image-level label and a pixel-level label, and correspondingly setting the established labels for the images in the anterior chamber angle data set; S4, constructing a deep neural network model fused with weak supervision metric learning; S5, training and verifying the constructed model by using the training set and the verification set to obtain an optimal network model; and S6, inputting the test set into the optimal network model to obtain anterior chamber angle image classification. According to the method, anterior chamber angle image classification can be quickly and accurately realized.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, and more specifically, to a method for grading anterior chamber angle images fused with weakly supervised metric learning. Background technique [0002] Glaucoma is a complex disease that causes nerve damage and eventually irreversible blindness if not diagnosed and treated early. Anterior chamber angle grading is one of the main means for doctors to evaluate glaucoma and formulate treatment plans. [0003] 1. Glaucoma detection. In recent years, with the widespread application of deep neural networks in medical image analysis, more and more deep neural network models have been designed for glaucoma detection. Most of these network models are used for anterior segment optical coherence tomography (Anterior Segment Optical Coherence Tomography, AS-OCT) or fundus images. For example, some researchers focused on the anterior chamber angle structure by segmenting the colle...

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
Patent Type & Authority Applications(China)
IPC IPC(8): G06V10/764G06V10/82G06V10/772G06K9/62G06T3/40G06T5/00G06T5/50G06T7/00G06T7/238G06N3/04G06N3/08
CPCG06T3/40G06T7/0012G06T7/238G06T5/50G06N3/08G06T2207/20016G06T2207/30041G06T2207/20221G06T2207/20084G06T2207/20081G06N3/045G06F18/24137G06F18/214G06T5/70
Inventor 贾西平黄静琪关立南聂栋崔怀林廖秀秀林智勇
Owner GUANGDONG POLYTECHNIC NORMAL 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