Pseudo CT synthesis method and application based on deep convolutional neural network

A deep convolution and neural network technology, applied in the fields of applications, radiological diagnostic instruments, medical science, etc., can solve problems affecting the prediction accuracy of pseudo-CT, achieve rapid generation and save time

Inactive Publication Date: 2019-11-19
SOUTHERN MEDICAL UNIVERSITY
View PDF5 Cites 18 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Based on the current statistical learning method to establish a mapping model between the MRI voxel intensity value and the CT gray value, but this model requires manual extraction of features, and the extracted features directly affect the accuracy of pseudo-CT prediction. This application provides A pseudo-CT synthesis method and application based on deep convolutional neural network

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
  • Pseudo CT synthesis method and application based on deep convolutional neural network
  • Pseudo CT synthesis method and application based on deep convolutional neural network
  • Pseudo CT synthesis method and application based on deep convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

preparation example Construction

[0047] see Figure 1~4 , the application provides a pseudo-CT synthesis method based on a deep convolutional neural network, the method comprising the steps of:

[0048] Step 1: Collect patient MRI images, CT images and radiotherapy planning information;

[0049] Step 2: the MRI image voxel size is adjusted to be consistent with the CT image voxel size, and then the MRI image and the CT image are carried out image registration;

[0050] Step 3: adopt deep convolutional neural network to set up the correlation model between the MRI image of described registration and CT image, and adopt the mapping relation between MRI voxel intensity value and CT gray value of learning deep convolutional neural network;

[0051] Step 4: Predict pseudo-CT images of patients using the association model.

[0052]Further, in the step 1, the MRI image and the CT image include the MRI image and the CT image of the same part, and the MRI image and the CT image are acquired within a specified time i...

Embodiment 1

[0087] In this example, the pseudo-CT of nasopharyngeal carcinoma patients is generated, and the MRI images and CT images of 45 nasopharyngeal carcinoma patients are selected for model training and testing.

[0088] First, use MATLAB software to process the MRI image, and adjust the voxel size of the MRI image to be consistent with the CT image voxel size. Then use the 3D slicer software to perform rigid registration of the MRI images and CT images of all patients, export the registered MRI images, import them into MATLAB, obtain the masks of all CT images, and use the masks to align all MRI images with CT images. The image background is uniformly set to 0. All MRI voxels and CT voxels intensity distributions are normalized to the range [-1,1]. Select 30 patients as the training set, and perform data enhancement on the MRI images and CT images in the training set. The data enhancement method selected in this example is horizontal / vertical flip and 180° rotation, and export th...

Embodiment 2

[0103] This embodiment is based on the pseudo-CT synthesis method based on the deep convolutional neural network described in the above-mentioned embodiments, and provides a specific application method of the pseudo-CT synthesis method. The specific steps are:

[0104]The pseudo CT generated by the 15 test sets in Example 1 was imported into the Eclipse planning system for dose recalculation, and the setting of planning execution parameters such as the incident field was kept consistent with the clinical actual plan, and the dose distribution of the pseudo CT was compared with the original planned dose distribution. The comparison mode of the dose distribution that the present embodiment takes is mean error (ME), mean absolute error (MAE), gamma evaluation passing rate (2mm / 2%, 3mm / 3%), as shown in table 2:

[0105]

[0106] Table 1 Comparison of the dose distribution based on pseudo-CT recalculation and the real planned dose distribution in the test set

[0107] The result...

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 pseudo CT synthesis method and application based on deep convolutional neural network. At present, a used model needs to extract features artificially, wherein the extractedfeatures directly affect the accuracy of pseudo CT prediction. The invention provides the pseudo CT synthesis method based on the deep convolutional neural network, wherein the pseudo CT synthesis method includes the steps: 1, collecting an MRI image, a CT image and radiotherapy plan information of a patient; 2, adjusting the voxel size of the MRI image to be consistent with the voxel size of theCT image, and then carrying out image registration on the MRI image and the CT image; 3, building a correlation model between the registered MRI image and the CT image by the deep convolutional neural network, and learning the mapping relationship between the MRI voxel intensity value and the CT gray value by the deep convolutional neural network; and 4, predicting a pseudo CT image of the patient by the correlation model. The synthetic pseudo CT is more accurate.

Description

technical field [0001] The present application belongs to the technical field of medical radiation therapy, and in particular relates to a pseudo-CT synthesis method and application based on a deep convolutional neural network. Background technique [0002] Magnetic Resonance Imaging (MRI) and Computer Tomography (Computer Tomography, CT) are medical images commonly used in clinical practice. MRI uses hydrogen proton resonance imaging in tissues, without ionizing radiation, and has good soft tissue contrast, which is convenient for delineating organs and target areas. However, the bones in the images acquired by conventional MRI sequences are not clear, and the MRI image intensity information reflects the tissue proton density and magnetic relaxation characteristics (ie, the inertia of the dipole moment), but does not reflect the electron density of the tissue, so the images cannot be used for Dose calculations for radiation therapy planning. The bones in CT images are ver...

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): A61B5/055A61B6/03
CPCA61B5/055A61B5/7264A61B6/032A61B6/5247
Inventor 宋婷亓孟科吴艾茜陆星宇贾启源郭芙彤刘裕良周凌宏
Owner SOUTHERN MEDICAL UNIVERSITY
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