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

Thyroid nodule recognition model training method and system based on parameter migration

A thyroid nodule and recognition model technology, applied in computational models, neural learning methods, character and pattern recognition, etc., can solve the problems of severe speckle noise, unfavorable for the promotion of deep learning, and high task difficulty

Pending Publication Date: 2021-09-21
SHANDONG NORMAL UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Benefits of technology

In this patented method called migration learning, we use various images from another field or medical problem that makes identifying specific areas difficult without training complex models with many parameters. By doing these techniques, our methods can help identify tissue abnormalities more accurately than existing approaches such as histopathology analysis alone.

Problems solved by technology

Technological Problem addressed by this patented describes how current methods involve collecting and processing non structural (non-ultrasound) or full-length (unconventional X ray), leading to increased computational requirements and storage needs during doctor visually interpreting sonography scans. Additionally, existing methodologies have limitations when trying to accurately diagnose certain types of cancer like Thyroadenoma because they require expensive computer hardware and extensive amounts of data from various sources.

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
  • Thyroid nodule recognition model training method and system based on parameter migration
  • Thyroid nodule recognition model training method and system based on parameter migration
  • Thyroid nodule recognition model training method and system based on parameter migration

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] This embodiment discloses a method for training a thyroid nodule recognition model based on parameter migration, such as figure 1 shown, including the following steps:

[0037] Step 1: Acquire the labeled source domain data and ultrasound images of thyroid nodules;

[0038] Specifically, open source images for training are obtained from the ImageNet open source dataset, categories are merged, and labels are added to be used as source domain data for training. In order to obtain the parameters that can extract the low-dimensional features of pictures, this example extracts pictures from the ImageNet open source dataset for category merging, and selects four categories of people, animals, indoors, and traffic to form a pre-training dataset to train picture classification tasks. The data set format is in jpg format.

[0039] Crawl ultrasound images of thyroid nodules from foreign open source medical image websites, trim and add labels, and finally obtain ultrasound image...

Embodiment 2

[0065] This embodiment provides a thyroid nodule recognition model training system based on parameter migration, including:

[0066] The training data acquisition module is configured to: acquire the source domain data and the ultrasound image of the thyroid nodule that have been marked;

[0067] The pre-training module is configured to: use the source domain data to pre-train the convolutional neural network to obtain pre-training model parameters;

[0068] The migration training module is configured to: increase or decrease the fully connected layer on the basis of the convolutional neural network, the reserved layer adopts the pre-trained model parameters, and uses the ultrasound picture of the thyroid nodule as input to perform the modified convolution The neural network is trained to obtain a thyroid nodule recognition model.

Embodiment 3

[0070] This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, the above-mentioned parameter-based migration and Steps in a locally supervised learning-based thyroid nodule diagnostic system.

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 thyroid nodule recognition model training method and system based on parameter migration. The method comprises the following steps: acquiring marked source domain data and a thyroid nodule ultrasonic image; pre-training a convolutional neural network by adopting the source domain data to obtain pre-training model parameters; and on the basis of the convolutional neural network, increasing and decreasing the fully connected layers, training the modified convolutional neural network by adopting pre-training model parameters for the reserved layers and taking a thyroid nodule ultrasonic picture as input to obtain a thyroid nodule recognition model for recognition and auxiliary diagnosis of thyroid nodules. The thought of transfer learning is introduced so that the problem that training is difficult due to the fact that thyroid nodules affect data target recognition difficulty and features are not obvious is solved.

Description

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

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
Owner SHANDONG NORMAL UNIV
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