Premature infant retinal image classification method and device based on attention mechanism

A technology of retina of premature infants and a classification method, which is applied in the field of retinal image classification of premature infants based on attention mechanism, can solve the problems of unbalanced distribution of disease characteristics and background of ROP images, inaccurate classification of ROP images, etc., and achieves improved recognition. and classification efficiency, improving classification performance, and high reliability.

Active Publication Date: 2020-06-09
SUZHOU UNIV
View PDF4 Cites 36 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to solve the deficiencies in the prior art, the present invention provides a method and device for classifying retinal images of premature infants based on the attention mechanism, which s

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
  • Premature infant retinal image classification method and device based on attention mechanism
  • Premature infant retinal image classification method and device based on attention mechanism
  • Premature infant retinal image classification method and device based on attention mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0040] A method for classifying retinal images of premature infants based on an attention mechanism, comprising steps:

[0041] Step 1, preprocessing the 2D retinal fundus image to be identified to obtain the preprocessed 2D retinal fundus image;

[0042] Preprocessing includes: normalization of downsampling and mean subtraction;

[0043] Such as figure 1 As shown, the schematic diagram of the two-dimensional retinal fundus image classified in the present invention: (a) normal image, (b) 1 stage, (c) 2 stage, (d) 3 stage, (e) 4 stage, (f) Stage 5, (g) Additional disease.

[0044] In order to prevent GPU memory overflow, the original 640×480 two-dimensional retinal fundus image is down-sampled to 320×240 by bilinear interpolation; in order to improve the contrast of the image, the image is standardized by subtracting the mean value, that is, using the original image The pixel value minus its average pixel value.

[0045] Step 2. Input the 2D retinal fundus image preprocesse...

Embodiment 2

[0059] An attention-based retinal image classification device for premature infants, including:

[0060] A preprocessing module, configured to preprocess the two-dimensional retinal fundus image to be identified, to obtain the preprocessed two-dimensional retinal fundus image;

[0061] The identification module is used to input the preprocessed two-dimensional retinal fundus image into the pre-trained deep attention network model, and the classification result of the output image identifies the ROP image of retinopathy of prematurity;

[0062] The deep attention network model is to add a complementary residual attention module and a channel attention SE module respectively after the third residual layer and the fourth residual layer of the original ResNet18 network.

[0063]Further, the complementary residual attention module includes: channel attention SE module, maximum pooling layer, average pooling layer, two-dimensional convolutional layer, sigmoid layer; SE module is use...

Embodiment 3

[0070] The inventive method is compared with prior art, and experimental result is as follows:

[0071] In order to quantitatively evaluate the performance of the present invention, 1443 (850 normal images and 593 ROP images) two-dimensional retinal fundus images from 100 subjects were subjected to the preprocessing, and four commonly used classification evaluation indicators were used to test , including Accuracy, Precision, Recall, and F1-score. The definitions of Accuracy, Precision, Recall, and F1-score are as follows:

[0072]

[0073]

[0074]

[0075]

[0076] Among them, TP, FP, TN, and FN represent true positive, false positive, true negative, and false negative, respectively, and P and R represent precision (Precision) and recall (Recall), respectively.

[0077] The present invention evaluates and compares the original ResNet18 network, the method of Zhang et al., and the deep attention network model described in the present invention in the test data se...

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 a premature infant retinal image classification method and device based on an attention mechanism, and the method comprises the steps: carrying out the preprocessing of a to-be-recognized two-dimensional retinal fundus image, and obtaining a preprocessed two-dimensional retinal fundus image; inputting the preprocessed two-dimensional retinal fundus image into a pre-traineddeep attention network model, and outputting the classification result of the image to identify a premature infant retinopathy ROP image, wherein the deep attention network model is formed by respectively adding a complementary residual attention module and a channel attention SE module behind a third residual layer and a fourth residual layer of the original ResNet18 network. According to the method, rich and important global and local information can be obtained, so that the network can learn correct lesion features, the classification network can better solve the problem of great data imbalance between lesions and backgrounds, and the classification performance of the deep attention network model is improved.

Description

technical field [0001] The invention relates to the technical field of retinal image classification, in particular to an attention mechanism-based retinal image classification method and device for premature infants. Background technique [0002] Retinopathy of Prematurity (ROP) is a retinal vascular proliferative disease, mainly seen in premature infants (gestational age less than 32 weeks) and low birth weight infants (weight less than 1500 grams) and high levels of oxygen supplementation, It is the leading cause of childhood blindness worldwide. Because the imaging quality is affected by factors such as focus, light, and eye movement, there are problems such as very low quality of retinal fundus images in premature infants, and unbalanced distribution of lesion areas and backgrounds. Therefore, there are great challenges in the automatic recognition of ROP based on retinal fundus color images. sex. [0003] In recent years, with the rapid development of deep learning, m...

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): G06K9/62G06N3/04
CPCG06V2201/03G06N3/045G06F18/214G06F18/241
Inventor 陈新建彭圆圆朱伟芳
Owner SUZHOU UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap