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

Focus segmentation model training and application method based on semi-supervised learning

A semi-supervised learning and segmentation model technology, applied in the field of image recognition, can solve the problems of difficulty in extracting features of brain MRI images, poor effect, etc., to achieve good model training effect, reduce noise, and save costs.

Active Publication Date: 2021-07-20
GUANGDONG UNIV OF TECH
View PDF9 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the technical problems of the poor training effect of the sample model of the brain MRI image described in the above-mentioned prior art and the difficulty in feature extraction of the complex brain MRI image, the present invention provides a lesion segmentation model training based on semi-supervised learning, The application method and technical scheme are as follows:

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
  • Focus segmentation model training and application method based on semi-supervised learning
  • Focus segmentation model training and application method based on semi-supervised learning
  • Focus segmentation model training and application method based on semi-supervised learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0044] like figure 1 As shown, the lesion segmentation model training method based on semi-supervised learning includes steps:

[0045] S1: Build a deep learning network model in the deep learning framework, obtain all brain MRI image training samples, divide the unlabeled brain MRI images into the unlabeled training set, and divide the brain MRI images that have been correctly labeled with the lesion area label and The corresponding labels are included in the labeled training set;

[0046] The label is the lesion area in the manually marked image.

[0047] S2: Preprocessing all brain MRI image training samples;

[0048] The preprocessing includes: one or more of image format conversion, acquisition time correction, head movement correction, image fusion between different imaging methods, image registration and spatial smoothing.

[0049] S3: Input the preprocessed brain MRI image training samples into the shared encoder;

[0050] S4: The shared encoder performs multiple d...

Embodiment 2

[0077] This embodiment provides a method for applying a lesion segmentation model based on semi-supervised learning, such as figure 2 shown, including steps:

[0078] S101: input the MRI image of the brain to be segmented into the trained deep learning network model;

[0079] S102: Preprocessing the brain MRI image to be segmented;

[0080] S103: Input the preprocessed data into the shared encoder;

[0081] S104: Perform multi-domain attention mechanism processing on the results of each downsampling of the shared encoder, and input the features obtained by the last downsampling to the main decoder;

[0082] S105: The main decoder upsamples the input data. After each upsampling, the main decoder first performs feature fusion on the result of this upsampling and the output features of the multi-domain attention mechanism processing, and then performs the next step. Upsampling, the features obtained by the last upsampling are fused with the corresponding features output by th...

Embodiment 3

[0085] This embodiment provides a lesion segmentation training model based on semi-supervised learning, the model structure is as follows image 3 As shown, during the training process, the model includes: a data preprocessing module, a shared encoder, a multi-domain attention mechanism module, a main decoder, a data interference module, an auxiliary decoder, and a calculation loss module.

[0086] Brain MRI image training samples include unlabeled brain MRI images and labeled brain MRI images, divide the unlabeled brain MRI images into the unlabeled training set, and divide the brain MRI images that have been correctly labeled with the lesion area label And the corresponding labels are classified into the labeled training set;

[0087] After the model receives the input training samples, it first inputs the training samples into the data preprocessing module, and the data preprocessing module preprocesses the input data; then the preprocessed data is input into the shared enc...

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 discloses a focus segmentation model training and application method based on semi-supervised learning, relates to the technical field of image recognition, and solves the technical problem that an Alzheimer's disease lesion is difficult to mark due to a complex brain structure and excessive noise. According to the focus segmentation model, a multi-domain attention mechanism is added when down-sampling and up-sampling are carried out on an image, focus area features needing to be segmented are emphasized, the attention degree of the model on lesion areas in the image is improved, non-lesion area features are inhibited, and therefore the feature characterization capacity of a feature map is improved; in addition, a semi-supervised learning method is added into the model, a network model is trained by using unlabeled data, and a relatively good model training effect is achieved by using a small amount of labeled data sets, so that the cost of manual labeling is greatly saved, and convenient auxiliary diagnosis is provided for clinic.

Description

technical field [0001] The present invention relates to the technical field of image recognition, and more specifically, to a lesion segmentation model training and application method based on semi-supervised learning. Background technique [0002] Alzheimer's disease (AD) is a clinical syndrome characterized by progressive degeneration of memory function and cognitive function. It is the main cause of dementia and a major cause of human death. AD has a long incubation period, and its clinical manifestations will gradually deteriorate over time. Medically, its clinical manifestations are divided into three stages: subjective cognitive impairment (Subjective Cognitive Impairment, SCI), mild cognitive impairment, and mild cognitive impairment. Disorder (Mild Cognitive Impairment, MCI) and AD. Among them, SCI belongs to a stage that normal human brain aging will pass through, and usually occurs in most people as they age. MCI is an intermediate state between normal aging and ...

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): G06T7/11
CPCG06T7/11G06T2207/10088G06T2207/20081G06T2207/30016G06T2207/20221Y02T10/40
Inventor 徐超王卓薇陈子洋陈立宜
Owner GUANGDONG UNIV OF TECH
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