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

Chest radiography multi-label classification method based on self-correcting label generation network

A classification method and multi-label technology, applied in the field of medical image processing, can solve problems such as unbalanced prediction effects, and achieve the effects of alleviating the unbalanced phenomenon of model prediction, improving performance, and improving prediction accuracy.

Active Publication Date: 2020-06-12
FUDAN UNIV
View PDF7 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The object of the present invention is to provide a chest radiograph multi-label classification method that can solve the unbalanced prediction effect phenomenon generated on the chest radiograph multi-label classification problem

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
  • Chest radiography multi-label classification method based on self-correcting label generation network
  • Chest radiography multi-label classification method based on self-correcting label generation network
  • Chest radiography multi-label classification method based on self-correcting label generation network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0036] It can be seen from the background technology that most of the previous research regards multi-label as an independent single-label problem and ignores the correlation between labels. This ultimately leads to an imbalance in the model's prediction performance, that is, it performs well on easier-to-predict disease labels, but performs poorly on harder-to-predict disease labels, such as pneumonia.

[0037]The present invention conducts further research on the above problems. The self-modifying label generation network (SLGN) provided in the present invention can simultaneously capture and utilize the spatial association, semantic association and co-occurrence between labels, thereby maximizing the improvement of the model. prediction accuracy. The present invention employs an encoder-decoder structure. In image captioning tasks, CNNs are used as encoders to extract image features. Image features are an abstract representation of images, including high-level semantic in...

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 belongs to the field of medical image processing, and particularly relates to a chest radiography multi-label classification method based on a self-correcting label generation network. According to the invention, a self-correcting label generation network model is constructed and is used for multi-label classification of chest radiography; the network comprises a ResNet image encoder, a self-correcting attention mechanism module and a GRU decoder, the image encoder is used for acquiring high-level semantic features of an input image, namely a chest radiography image, generating image feature representation and outputting the image feature representation to the self-correcting attention mechanism module; the self-correcting attention mechanism module generates an attention mapcorresponding to the current time step at each moment according to the state information and the image features output at the previous moment, and outputs a context feature vector to the decoder; andthe decoder generates a label corresponding to the image at the current moment according to the context feature vector and a label word vector generated at the previous moment. According to the method, the problem of unbalanced prediction effect in chest radiography multi-label classification is effectively solved.

Description

technical field [0001] The invention belongs to the field of medical image processing, and in particular relates to a multi-label classification method for chest radiographs. Background technique [0002] Chest radiography is widely used in clinical diagnosis and treatment, and is one of the main means of detecting and diagnosing chest abnormalities. Specialized physicians read the chest radiographs and wrote text reports explaining the findings. This is a job that requires rich medical experience and high concentration. For inexperienced doctors, it is easy to make mistakes in diagnosing medical images. And for doctors in densely populated areas, reading medical images is time-consuming and tedious. Therefore, it is very important to study computer-aided diagnosis and treatment algorithms to help doctors better diagnose medical images. [0003] In general, chest radiographs usually contain one or more disease labels, so this is an image multi-label classification problem...

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): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/241G06F18/214
Inventor 张玥杰胡玥琳
Owner FUDAN 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