Diabetic retinopathy classification method by using super lightweight SqueezeNet network

A diabetic retina and ultra-lightweight technology, applied in the field of deep learning, can solve problems such as the inability to make full use of medical image information, the accuracy cannot be improved, and the accuracy is limited, so as to improve the accuracy and reliability of classification and reduce the number of models The effect of the parameter

Inactive Publication Date: 2018-09-14
NORTHEASTERN UNIV
View PDF5 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The shortcomings of the above grading methods are: first, the manually defined features have limitations, and cannot make full use of the information in medical

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
  • Diabetic retinopathy classification method by using super lightweight SqueezeNet network
  • Diabetic retinopathy classification method by using super lightweight SqueezeNet network
  • Diabetic retinopathy classification method by using super lightweight SqueezeNet network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0036] Embodiment: a kind of diabetic retinopathy classification grading method (as figure 1 The flow chart of the method disclosed in the present invention), the core of the classification and grading method is: preparing a large number of SLO fundus photos for each type of diabetic retinopathy and performing preprocessing and data amplification; Convolutional neural network; based on a large number of ophthalmoscope photos, the deep convolutional neural network is trained, so that the final output value of the deep convolutional neural network conforms to the grading result of the ophthalmoscope photo; thus, the trained deep convolutional neural network can be used to automatically Perform disease classification.

[0037] The diabetic retinopathy classification and grading method using the ultra-lightweight SqueezeNet network includes the following steps:

[0038]1) Prepare a photo library, which contains several ophthalmoscope photos including diagnostic markers, and each ...

example 2，3，3.0，； pic 8(c) example 3

[0050] Figure 8 is an example of the four images and scores judged after normalization to probability. Figure 8(a) is example 1, the real label is class 0, the predicted probability of class 0 is 0.993, and the judgment is correct; Figure 8(b) is example 2, the real label is class 3, the predicted probability of class 3 is .0, and the judgment is correct ; Figure 8(c) is example 3, the real label is class 2, the predicted probability of class 1 is 0.562, and the judgment is wrong; Figure 8(d) is example 4, the real label is class 1, the predicted probability of class 1 is 0.613, and the judgment is correct.

[0051] Introduce the functional definitions of several main structures of the ultra-lightweight SqueezeNet network structure used in the present invention: In functional analysis, convolution, convolution or convolution are generated by two functions f and g to generate a third function A mathematical operator that characterizes the area of ​​the overlapping part of the f...

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 diabetic retinopathy classification method by using a super lightweight SqueezeNet network. The method comprises the following steps: preparing lots of SLO (Scanning Laser Ophthalmoscope) fundus photographs aiming at each type of diabetic retinopathy and performing preprocessing and data amplification; establishing a super lightweight SqueezeNet deep convolutional neuralnetwork containing a fire module; training the deep convolutional neural network based on the lots of fundus photographs, and enabling the final output value of the deep convolutional neural network to accord with the classification results of the fundus photographs; automatically carrying out disease classification by utilizing the trained deep convolutional neural network. According to the method disclosed by the invention, due to application of the lots of fundus photographs comprising diagnostic markers, operations of automatically learning needed features from a training case library andperforming classification judgment are realized by virtue of super lightweight deep learning network and a few parameters, and data features for judgment and deep convolutional neural network parameters are continuously corrected in the training process, so that the classification accuracy and reliability in realistic application scenarios can be greatly improved.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method for grading each type of diabetic retinopathy. Background technique [0002] Diabetes is an endocrine disease that seriously affects human health, and its disability and mortality rate is second only to cardiovascular and cerebrovascular diseases and cancer. The disease not only brings great suffering to human beings, but also brings many complications, among which diabetic retinopathy (referred to as "diabetic retinopathy", Diabetic retinopathy, DR) has the highest incidence rate and the greatest impact on vision. Because every diabetic patient has the possibility of developing DR, and DR is progressive and irreversible. Therefore, how to accurately screen diabetic patients without obvious visual impairment for the presence of DR not only provides an opportunity for early diagnosis and early treatment to save the patient's visual function, but also saves a lot of...

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
CPCG06N3/045G06F18/24G06F18/214
Inventor 陈大力梅丹蕾朱姗姗王孝阳罗凌佟萌萌
Owner NORTHEASTERN 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