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

Method of image segmentation using CN

A technology for inputting images and feature maps, applied in the field of image segmentation using CNN, which can solve problems such as large additional computing costs

Active Publication Date: 2020-07-03
HONG KONG APPLIED SCI & TECH RES INST
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Most extensions and modifications to U-Net are made to improve accuracy, but this results in substantial additional computational cost

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
  • Method of image segmentation using CN
  • Method of image segmentation using CN
  • Method of image segmentation using CN

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0046] A CNN is a neural network with multiple hidden layers, at least some of which are convolutional layers, where each convolutional layer is used to perform a convolution or dot product with the input provided to that layer. The CNN is implemented and operated by a computing device programmed with program codes for performing data operations according to the network structure of the CNN.

[0047] This paper discloses a CNN model that improves medical image segmentation performance relative to U-Net while maintaining or reducing computational requirements compared to U-Net. In particular, this CNN model employs a multi-scale context aggregation module to achieve the first goal of improving segmentation performance, and a CFS module to reduce computation requirements.

[0048] Before elaborating on the disclosed CNN model, an overview of U-Net is provided below. U-Net was proposed by O.RONNEBERGER, P.FISCHER and T.BROX in 2015 in "U-Net: Convolutional Networks for Biomedica...

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

In the convolutional neural network (CNN) that uses the encoder-decoder structure for image segmentation, the multi-scale context aggregation module receives the final encoded feature map from the encoder, and sequentially aggregates the feature maps from the global scale to the local scale. Scale context to strengthen the semantic relationship of different scale contexts, thereby improving segmentation accuracy. Multi-scale context is obtained by calculating the hole convolution with different expansion rates on the feature map. In order to reduce the amount of calculation required for CNN, achannel-based feature selection (CFS) module is used in the decoder to merge the two input feature maps. Each feature map is processed by a global pooling layer, followed by a fully connected layer or a 1*1 convolutional layer to select highly active channels. Through subsequent multiplication by channel and addition by element, only the channels with high activation in the two feature maps are retained and enhanced in the merged feature map.

Description

[0001] Abbreviation list [0002] 2D two-dimensional [0003] 3D three-dimensional [0004] CFS channel-based feature selection [0005] CNN convolutional neural network [0006] CPU central processing unit [0007] CT Computed Tomography [0008] DS DICE score [0009] FLOP floating point operation [0010] GP global pooling [0011] GPU graphics processing unit [0012] MB megabyte [0013] MRI magnetic resonance imaging [0014] OCT optical coherence tomography [0015] PET Positron Emission Tomography [0016] TDP thermal design power technical field [0017] This disclosure generally relates to image segmentation using CNNs. In particular, the present disclosure relates to the network structure of CNNs for improving segmentation accuracy and reducing computational requirements. Background technique [0018] Medical image segmentation involves the extraction of anatomical regions of interest from a medical image or series of images through an automatic or s...

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): G06T7/10G06N3/04G06N3/08
CPCG06T7/10G06T2207/20081G06T2207/20084
Inventor 刘尚平王陆张平平卢湖川
Owner HONG KONG APPLIED SCI & TECH RES INST
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