General endoscope image enhancement method and device based on generative adversarial neural network

A neural network and image enhancement technology, applied in the field of medical image processing, can solve problems such as poor robustness, lack of versatility, and no matching public data sets, and achieve the effect of improving accuracy and strong adaptability

Pending Publication Date: 2020-10-16
BEIJING INSTITUTE OF TECHNOLOGYGY
View PDF5 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Endoscopic images involve difficulty in obtaining patient privacy, the amount of available data is much smaller than that of natural images, and there is no paired public data set available
Therefore, these methods can only train the model with simulated data, and the performance on real endoscopic images is not optimistic
And each different method is only for one kind of image quality problem, lack of versatility
Therefore, the existing endoscopic image quality enhancement methods using paired simulated data for training are less robust to real endoscopic scenes, and different methods are only applicable to specific endoscopic image quality defects and are not universal. sex

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
  • General endoscope image enhancement method and device based on generative adversarial neural network
  • General endoscope image enhancement method and device based on generative adversarial neural network
  • General endoscope image enhancement method and device based on generative adversarial neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0018] Such as image 3 As shown, this general endoscopic image enhancement method based on generating an adversarial neural network includes the following steps:

[0019] (1) Use a generative adversarial network to train a single model for image quality defects that exist in most endoscopic images;

[0020] (2) Extract part of the network layer of the generator in the model as a general pre-training model;

[0021] (3) For different endoscope image quality defects, train on the basis of the pre-trained model in step (2), and finally obtain a neural network model with better performance for different endoscope image quality defects.

[0022] The present invention trains a single model aiming at the image quality defect existing in most endoscopic images by using the generative confrontation network, which is more adaptable to the real endoscopic scene than the model trained by using paired simulated data. It is strong and can better meet the actual needs of doctors in surgic...

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 general endoscope image enhancement method and device based on the generative adversarial neural network, the requirement of a task can be met, the adaptability to a real scene is high, the method and device can be applied to different endoscope image quality enhancement problems, and the training precision is greatly improved. The method comprises the steps of: (1)using a generative adversarial network for training a single model for image quality defects existing in most endoscope images; (2) extracting a part of network layer of a generator in the model as a universal pre-training model; and (3) aiming at different endoscope image quality defects, carrying out training on the basis of the pre-training model in the step (2), and finally obtaining a neural network modelwith better performance aiming at different endoscope image quality defects.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, in particular to a general endoscopic image enhancement method based on generating adversarial neural networks and a general endoscopic image enhancing device based on generating adversarial neural networks. Background technique [0002] An endoscope, consisting of a light source, a lens, and a bendable part, is one of the most commonly used medical devices. During endoscopic imaging, different image quality issues arise. Due to the narrow cavity in the body, uniform full-field illumination cannot be obtained, and the obtained endoscopic image is uneven in brightness; due to the moist and smooth surface of the mucous membrane in the body, the obtained endoscopic image is full of bright spots; due to intraoperative resection The diseased tissue will produce a large amount of white smoke, and the obtained endoscopic image will be blurred. In operations using laser therapy...

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): G06T5/00G06N3/04G06N3/08
CPCG06N3/08G06T2207/10068G06T2207/20081G06T2207/20084G06N3/045G06T5/77
Inventor范敬凡杨健王涌天李雅婷
OwnerBEIJING INSTITUTE OF TECHNOLOGYGY