Unlock instant, AI-driven research and patent intelligence for your innovation.

Thyroid nodule property classification method based on SPECT image

A technology of thyroid nodule and classification method, applied in the field of image processing, can solve the problems of balanced image quantity, difficult to achieve good effect, large amount of data, etc., and achieve the effect of high automatic classification, cost saving and high accuracy rate

Active Publication Date: 2021-12-24
HARBIN INST OF TECH AT WEIHAI
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the amount of data required by the deep learning method is large, and the amount of data of various types is required to be balanced.
There are the following problems in classifying and judging the nature of thyroid nodules by SEPCT images: first, the acquisition cost is high, and a large number of samples cannot be obtained; quality image quantity balance
Therefore, the existing deep learning methods are difficult to achieve good results

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 property classification method based on SPECT image
  • Thyroid nodule property classification method based on SPECT image
  • Thyroid nodule property classification method based on SPECT image

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0048] In order to solve the problems existing in the prior art, the present invention provides a method for classifying the properties of thyroid nodules based on SPECT images, which realizes automatic classification with high accuracy by mining difficult samples and performing metric learning through Siamese networks.

[0049]The above twin network is actually a dual network composed of two identical networks. The two networks perform metric learning by sharing and updating the same weights, learn from fewer samples, and map the distance between different data to a metric space. Metric learning methods aim to learn task-specific distance or similarity functions by automatically constructing metrics for labeled data. Compared with traditional supervised learning, better generalization ability makes it more suitable for small data problems. In addition, metric learning focuses on the relationship between each data rather than each category, and this class independence improve...

Embodiment 2

[0082] A SPECT image dataset is obtained, which consists of a total of 273 SPECT images of four types of thyroid nodules. Among them, there were 30 sheets of cold nodules, 199 sheets of cool nodules, 10 sheets of warm nodules, and 34 sheets of hot nodules. The category of thyroid nodules in the samples comes from the conclusion of clinical diagnosis.

[0083] When using the data set training model, divide random training set and test set according to 8:2, use the training set to train according to the method of 5-fold crossover, use the test set to calculate the performance index of the classification method of the present invention, and evaluate the performance of the classification method. Repeat the above process 5 times. According to the classification results, the performance indicators are compared with the real results.

[0084] Adjust the thyroid SPECT image to a 224×224 thyroid SPECT image. The classification convolutional neural network uses the ResNet50 network, ...

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 property classification method based on an SPECT image, and the method comprises the steps: obtaining an SPECT image data set of a thyroid, inputting the SPECT image data set of the thyroid into a thyroid nodule property classification model for training, and obtaining a trained thyroid nodule property classification model, wherein the network structure of the thyroid nodule property classification model comprises a classification convolutional neural network and a twin neural network; and inputting a to-be-tested SPECT image based on the trained thyroid nodule property classification model to obtain a thyroid nodule property classification result. According to the invention, through a unique network structure and mining of difficult samples, a qualified classification model is trained with few training samples, so that the cost is saved, the efficiency is improved, and high-accuracy automatic classification is realized.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for classifying properties of thyroid nodules based on SPECT images. Background technique [0002] Single-photon emission computed tomography (Single-Photon Emission Computed Tomography, SPECT) is an important nuclear medicine imaging technique. [0003] SPECT images can be used to classify nodules into cold, cool, warm, and hot nodules according to the different nuclide uptake capabilities of nodules, which can assist clinical judgment of the nature of thyroid nodules. Hot nodules indicate that the imaging concentration of the contrast agent in the thyroid nodule is higher than that of the surrounding normal tissue; warm nodules indicate that the imaging concentration of the contrast agent in the thyroid nodule is the same as that of the surrounding normal thyroid tissue; cold nodules indicate that the imaging concentration of the contrast agent in the thyroid nodule is ...

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): G06K9/62G06K9/46G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/22G06F18/24
Inventor 马立勇王联芳张湧孙明健
Owner HARBIN INST OF TECH AT WEIHAI