Medical information cross-modal hash coding learning method based on generative adversarial network

A hash coding, medical information technology, applied in character and pattern recognition, image analysis, image enhancement and other directions, can solve the problem of not making full use of semantic information, unable to achieve semantic association and so on

Active Publication Date: 2020-05-08
KUNMING UNIV OF SCI & TECH
View PDF5 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Both text-based and content-based medical image retrieval are single-modal retrieval, which can only rely on the semantic information or even annotation information of a single modality for retrieval between modal data, and cannot make full us...

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
  • Medical information cross-modal hash coding learning method based on generative adversarial network
  • Medical information cross-modal hash coding learning method based on generative adversarial network
  • Medical information cross-modal hash coding learning method based on generative adversarial network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0043] Embodiment 1: as Figure 1-4 As shown, a method for learning cross-modal hash coding of medical information based on generative confrontation network, the specific steps of the method are as follows:

[0044] Step1. Perform feature extraction on the chest CT image-text data; first perform chest CT image preprocessing, cut out the ROI image block, and then extract images from the ROI image block and chest CT image-text data through the CMSFF model and the bag-of-words model features and textual features;

[0045] Further, the specific steps of the Step1 are as follows:

[0046] Step1.1, firstly perform data preprocessing; for the image data set, in order to avoid the loss of pixels caused by direct compression from the size of 512*512 to 224*224, the method of cutting the original CT image is adopted; according to the lung nodules on the slice The diameters are different, and the Region Of Interest, ROI image block with the size R={16*16, 32*32, 64*64, 128*128} is cut ...

Embodiment 2

[0080] Embodiment 2: as Figure 1-5 As shown, a kind of medical information cross-modal hash coding learning method based on generative confrontation network, this embodiment is the same as embodiment 1, the difference is:

[0081] Among them, Step6, the hash codes of different modalities learned through Step5 are stored in the hash code database, and at the same time, a trained model is obtained for the cross-modal retrieval system. Input the text data of any group of pulmonary nodules, obtain the corresponding hash code through cross-modal retrieval, and then retrieve the optimal result in the hash code database. The retrieval process is as follows: figure 2 shown.

[0082] The concrete steps of described Step6 are:

[0083] For the text mode, let its input be y, and learn its hash code through the GANHL model, such as the formula C y = h (y) (f (y) (y; θ t , θ D )) shown.

[0084] The query data of the given text is realized through the GANHL retrieval model, and t...

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 a medical information cross-modal hash coding learning method based on a generative adversarial network, and belongs to the technical field of medical information processing and information retrieval. According to the method, a generative adversarial network is adopted to learn hash codes of chest CT images and texts, and the learned hash codes are constrained through a semantic similarity matrix. Finally, accurate hash codes are learned, and semantic association between the two modes is successfully achieved. According to the method, on the basis of single-layer fine-grained pulmonary nodule features, more complete feature information of the three-dimensional pulmonary nodule is extracted, and a hash code generation model is obtained by adopting a supervised training mode in the method, so that relatively high accuracy is realized in cross-modal retrieval.

Description

technical field [0001] The invention relates to a method for learning cross-modal hash coding of medical information based on a generative confrontation network, and belongs to the technical field of medical information processing and information retrieval. Background technique [0002] Research on computer-aided diagnosis through deep learning to solve some problems in the medical field has attracted more and more attention from researchers and doctors, and lung cancer is currently one of the most widely studied diseases. Radiologists detect early lung cancer by screening nodules on chest CT images, and save the detection results in text form as the basis for clinicians' diagnosis. In the early days, the diagnosis of the malignancy of pulmonary nodules was mainly by setting thresholds, observing the changes in nodule volume at different times, and finally using a standard formula to evaluate the growth rate of nodules. At present, scholars have carried out multi-modal retr...

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): G06T7/00G06K9/62
CPCG06T7/0012G06T2207/20104G06T2207/10081G06T2207/30064G06F18/22
Inventor 黄青松贺周雨赵晓乐刘利军冯旭鹏
Owner KUNMING UNIV OF SCI & TECH
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