Ultrasonic and nuclear magnetic image registration method and device based on multi-scale supervised learning

A supervised learning and nuclear magnetic image technology, applied in the field of medical image processing, can solve the problems of sparse artificial labels and inability to carry out effective supervision, and achieve the effect of improving speed and accuracy

Pending Publication Date: 2020-05-01
BEIJING INSTITUTE OF TECHNOLOGYGY
View PDF5 Cites 28 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] 1. Manual labeling, errors cannot be avoided
[0008] 2. Artificial labels are sparse, and some areas cannot be effectively supervised

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
  • Ultrasonic and nuclear magnetic image registration method and device based on multi-scale supervised learning
  • Ultrasonic and nuclear magnetic image registration method and device based on multi-scale supervised learning
  • Ultrasonic and nuclear magnetic image registration method and device based on multi-scale supervised learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] For multimodal image registration based on deep learning, the biggest challenge is that it is difficult to obtain a gold standard. Due to the low quality of ultrasound images, there is much noise. Traditional algorithms are also difficult to obtain satisfactory registration results. The applicant segmented the corresponding important organ structures in the ultrasound and magnetic resonance images, matched the difference with the segmentation label, the gradient descriptor-based similarity measure difference between the deformed ultrasound image and the magnetic resonance image, and smoothed the deformation field The regular term, these three loss functions together serve as the training loss of the neural network. Important anatomical structures that need to be segmented include organ contours, boundaries of tubular structures such as blood vessels. In order to pursue higher registration accuracy and improve the robustness of the registration network, the traditional...

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 an ultrasonic and nuclear magnetic image registration method and device based on multi-scale supervised learning, a deformation field from ultrasonic to magnetic resonance can be predicted only by inputting ultrasonic and magnetic resonance images in a test stage, registration from ultrasonic to magnetic resonance is completed, iteration of a traditional algorithm is not needed, and the speed and accuracy are greatly improved. The method comprises the following steps: (1) collecting liver three-dimensional magnetic resonance images of a plurality of patients and ultrasonic imagescorresponding to the liver three-dimensional magnetic resonance images as training samples, and segmenting important tissue regions as label images; (2) constructing a registration basic framework based on 3DUnet; (3) obtaining a neural network based on multi-scale supervision; (4) rigidly registering the corresponding magnetic resonance and ultrasonic images; inputting the processed magnetic resonance and ultrasonic images as two channels into a neural network, and training the neural network by using three loss functions of MIND similarity measurement, segmentation label loss and deformationfield smoothing term; and (5) registering the magnetic resonance images and the ultrasonic images according to the registration parameters.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, in particular to a multi-scale supervised learning-based ultrasound and nuclear magnetic image registration method, and a multi-scale supervised learning-based ultrasound and nuclear magnetic image registration device. Background technique [0002] With its high incidence and high mortality, liver cancer has become one of the major disease threats facing the people of our country. Real-time guided liver tumor ablation surgery based on ultrasound US (ie, ultrasound scan, ultrasound scan) has become a key development direction of clinical medicine due to its advantages of low cost and small wound. However, compared with MR (Magnetic Resonance Imaging, Magnetic Resonance Imaging) images that can only be acquired before surgery, ultrasound images have low imaging quality and narrow field of view, making it difficult to provide comprehensive anatomical details. Therefore, in ...

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): G06T7/33G06T7/10G06K9/62G06N3/04G06N3/08
CPCG06T7/33G06T7/10G06T2207/10088G06T2207/10132G06T2207/20081G06T2207/20084G06T2207/30056G06T2207/30096G06T2207/30081G06N3/08G06V10/751G06N3/045
Inventor 杨健范敬凡王涌天邓巧玲
Owner BEIJING INSTITUTE OF TECHNOLOGYGY
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