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

CT image data automatic classification method and device based on CNN and GAN

A CT image, automatic classification technology, applied in the field of data processing, can solve the problem that radiologists are difficult to distinguish sub-centimeter pulmonary nodules, etc., to achieve strong market application prospects and the effect of promoting daily work

Pending Publication Date: 2020-07-14
刘雷 +2
View PDF6 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although computed tomography (CT) examination is widely used in practice, it is still difficult for radiologists to distinguish different types of subcentimeter pulmonary nodules

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
  • CT image data automatic classification method and device based on CNN and GAN
  • CT image data automatic classification method and device based on CNN and GAN
  • CT image data automatic classification method and device based on CNN and GAN

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] This embodiment discloses as figure 1 Shown a kind of CT image data automatic classification method based on CNN and GAN, described method comprises the following steps:

[0035] S1 acquires CT image data to be classified;

[0036] S2 selects the image of the nodule itself to perform data enhancement processing to obtain a public extended data set;

[0037] S3 uses GAN to obtain the generation network and discrimination network for the public extended data set, and simultaneously trains to obtain the GAN synthetic data set;

[0038] S4 uses the CNN network to classify the GAN synthetic data set to obtain the final image data set.

[0039] In step S2, only the image of the nodule itself is used, and the nodule area in each CT scan is calculated according to the annotation of the radiologist, only the three CTs with the largest nodule area are selected, and the nodule-centered image is cut. 64×64 pixel image and named as the original dataset.

[0040] The specific dat...

Embodiment 2

[0051] This embodiment discloses a data processing procedure, in which only the image of the nodule itself is used. The nodule area in each CT scan was calculated based on the radiologist's annotations, and only the three CTs with the largest nodule area were selected. A 64 × 64 pixel image centered on the nodule was cut and named as the original dataset.

[0052] Sufficient data is important because even small CNNs contain thousands of parameters and are prone to overfitting. To avoid this problem, a common strategy is data augmentation. Common augmentation techniques include translation, rotation, scaling and flipping. The present invention first randomly translates the image pixel by pixel. Then rotate with the nodule as the center. Afterwards, the image is rescaled with a random ratio of 80% to 120%. Finally, the nodule patch is flipped upside down and sideways. The image generated by the above operation is 64 × 64 pixels, which is consistent with the original datase...

Embodiment 3

[0059] This embodiment discloses a GAN structure, a gradually growing wGAN structure for generating images of lung adenocarcinoma. The generator consists of nine convolutional layers. First, 512-dimensional random noise is input into a fully connected layer, and then a 4×4 pixel feature map is generated by the first convolutional layer. Then, the feature maps are passed through four modules consisting of two convolutional layers. The detailed structure of the blocks is shown in Table 4. These modules continuously double the height and width of the feature map, and finally generate a 64 × 64 pixel image.

[0060] The discriminator also includes nine convolutional layers. It mirrors the generator, starting with a 64×64 pixel image, passing through four blocks consisting of two convolutional layers, and ending with a 4×4 pixel feature map. Finally, a mini-batch discriminative technique is added to the convolutional layer to make the training more stable. Finally, the feature ...

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 relates to the technical field of data processing, in particular to a CT image data automatic classification method and device based on CNN and GAN, and the method comprises the following steps: S1, obtaining CT image data to be classified; S2, selecting an image of the nodule to carry out data enhancement processing to obtain a public expansion data set; S3, obtaining a generation network and an identification network for the public expansion data set by using the GAN, and performing training at the same time to obtain a GAN synthesis data set; and S4, classifying the GAN synthetic data set by using a CNN network to obtain a final image data set. According to the method, the problem that most of existing researches about lung adenocarcinoma classification focus on radiomicsfeature modeling and other manual marking features, which are based on manual labeling, thus more burden problems are brought to doctors can be solved; according to the invention, the lightweight CNNmodel is also convenient to arrange in a hospital diagnosis system, daily work of radiologists is facilitated, development of precision medical treatment is promoted, and the invention has a very strong market application prospect.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a method and device for automatically classifying CT image data based on CNN and GAN. Background technique [0002] Efficient and accurate diagnosis of lung adenocarcinoma before surgery is of great significance to clinicians. Although computed tomography (CT) examination is widely used in practice, it is still difficult for radiologists to distinguish different types of subcentimeter pulmonary nodules. In this paper, the present invention proposes an automatic classification system for subcentimeter lung adenocarcinoma, which combines convolutional neural network (CNN) and generative adversarial network (GAN). The system is based on 2D nodule center CT patches for processing without manual labeling of information. A total of 206 postoperative pathologically marked nodules were analyzed, of which 30 were adenocarcinoma in situ (AIS), 119 were minimally invasive adenocar...

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): G16H30/40G16H50/20G06T7/00G06T7/62G06K9/62A61B6/03A61B6/00
CPCG16H30/40G16H50/20G06T7/0012G06T7/62A61B6/032A61B6/5205A61B6/5217G06T2207/30064G06F18/241
Inventor 刘雷周凌霄王云鹏
Owner 刘雷
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