Granted Patent B2

US 12,614,339 · App. 19/023,894 · Granted Apr 28, 2026

Inventors: Ping Chen (Shanxi, CN); Zhiqing Wei (Shanxi, CN); Sukai Wang (Shanxi, CN); Jiaotong Wei (Shanxi, CN); Xiaojie Zhao (Shanxi, CN); Yangxu Wu (Shanxi, CN)

Assignee: NORTH UNIVERSITY OF CHINA

G06T15/08G06T3/40G06V10/44G16H30/20G16H30/40

Abstract

Claims (86)

1 . A three-dimensional computed tomography (CT) imaging method based on X-ray dual projection, comprising:

collecting two digital radiography (DR) images with a perpendicular relationship for an object to be imaged;

obtain a three-dimensional volume of the object to be imaged corresponding to the two DR images with a perpendicular relationship based on a preset three-dimensional volume reconstruction model; and

slicing the three-dimensional volume to obtain three-dimensional CT imaging of the object to be imaged,

wherein, generating the preset three-dimensional volume reconstruction model comprises:

obtaining a training data set; wherein the training data set is composed of DR images of reference objects collected circumferentially;

training to-be-trained three-dimensional volume reconstruction model using DR images with a perpendicular relationship in the training data set to obtain the preset three-dimensional volume reconstruction model;

wherein, when obtaining the training data set, the method further comprises:

obtaining a reference three-dimensional volume and a reference pseudo-sinogram of each of the reference objects based on the training data set; wherein the reference pseudo-sinogram is obtained by stacking collected DR images of a same reference object into a three-dimensional matrix according to a collection order, summing sinograms in each layer column by column, and normalizing a sum result;

dividing the DR images of each reference object in the training data set into a plurality of DR image groups, and each of the DR image groups comprises two DR images with a perpendicular relationship;

wherein the training to-be-trained three-dimensional volume reconstruction model using DR images with a perpendicular relationship in the training data set to obtain the preset three-dimensional volume reconstruction model comprises:

inputting the plurality of DR image groups into the to-be-trained three-dimensional volume reconstruction model, training the to-be-trained three-dimensional volume reconstruction model, and respectively obtaining a training pseudo-sinogram and a training three-dimensional volume;

calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram and the reference three-dimensional volume according to a training order;

terminating the training and obtaining the preset three-dimensional volume reconstruction model in a case that any calculated loss value is less than a target loss value, or the number of training times reaches a target number of times.

2 . The method according to claim 1 wherein the to-be-trained three-dimensional 1 volume reconstruction model comprises an enhancement network model and a reconstruction network model;

the enhancement network model is a model based on an encoder-decoder network architecture, and configured to input a DR image group with a perpendicular relationship and output a corresponding training pseudo-sinogram; and

the reconstruction network model is an autoencoder network model for two-dimensional to three-dimensional volume mapping and configured to input the training pseudo-sinogram and output a corresponding training three-dimensional volume.

3 . The method according to claim 2 , wherein the enhancement network model to input a DR image group with a perpendicular relationship, and output a corresponding training pseudo-sinogram, comprises:

performing pattern and feature

extraction on the inputted DR image group, and performing a cascade operation on a results of the extraction;

performing latent space feature encoding on features after the cascade operation to obtain encoded latent space features; and

decoding the encoded latent space features to obtain the training pseudo-sinogram.

4 . The method according to claim 2 , wherein the reconstruction network model to input the training pseudo-sinogram, and output a corresponding training three-dimensional volume, comprises:

performing feature capture and enhancement on the input training pseudo-sinogram;

encoding a result of feature enhancement, and continuously reducing the spatial resolution on the basis of keeping the number of channels unchanged; and

decoding a result of the encoding, and continuously restoring the spatial resolution to a spatial resolution of the reference three-dimensional volume to obtain the training three-dimensional volume.

5 . The method according to claim 1 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

6 . The method according to claim 2 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

7 . The method according to claim 3 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises;

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

8 . The method according to claim 4 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function: the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

9 . A three-dimensional computer tomography (CT) imaging device based on X-ray dual projection, comprising:

a storage unit, configured to store a preset three-dimensional volume reconstruction model; generating the preset three-dimensional volume reconstruction model comprises: obtaining a training data set; wherein the training data set is composed of digital radiography (DR) images of reference objects collected circumferentially; training to-be-trained three-dimensional volume reconstruction model using DR images with a perpendicular relationship in the training data set to obtain the preset three-dimensional volume reconstruction model;

an acquisition unit, configured to collect two DR images with a perpendicular relationship for an object to be imaged;

an obtaining unit, configured to obtain a three-dimensional volume of the object to be imaged corresponding to the two DR images with a perpendicular relationship based on the preset three-dimensional volume reconstruction model; and

an imaging unit, configured to slice the three-dimensional volumes to obtain three-dimensional CT imaging of the object to be imaged,

wherein, generating the three-dimensional volume reconstruction model stored in the storage unit comprises:

when obtaining the training data set, obtaining a reference three-dimensional volume and a reference pseudo-sinogram of each of the reference objects based on the training data set; wherein the reference pseudo-sinogram is obtained by stacking collected DR images of a same reference object into a three-dimensional matrix according to a collection order, summing sinograms in each layer column by column, and normalizing a sum result;

dividing the DR images of each reference object in the training data set into a plurality of DR image groups, and each of the DR image groups comprises two DR images with a perpendicular relationship;

inputting the plurality of DR image groups into the to-be-trained three-dimensional volume reconstruction model, training the to-be-trained three-dimensional volume reconstruction model, and respectively obtaining a training pseudo-sinogram and a training three-dimensional volume;

calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram and the reference three-dimensional volume according to a training sequence;

terminating the training and obtaining the preset three-dimensional volume reconstruction model in a case that any calculated loss value is less than a target loss value, or the number of training times reaches a target number of times.

10 . An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, the processor is configured to execute the computer program to implement a three-dimensional computed tomography (CT) imaging method based on X-ray dual projection, the method comprises:

collecting two digital radiography (DR) images with a perpendicular relationship for an object to be imaged;

obtain a three-dimensional volume of the object to be imaged corresponding to the two DR images with a perpendicular relationship based on a preset three-dimensional volume reconstruction model; and

slicing the three-dimensional volume to obtain three-dimensional CT imaging of the object to be imaged,

wherein, generating the preset three-dimensional volume reconstruction model comprises:

obtaining a training data set; wherein the training data set is composed of DR images of reference objects collected circumferentially;

training to-be-trained three-dimensional volume reconstruction model using DR images with a perpendicular relationship in the training data set to obtain the preset three-dimensional volume reconstruction model;

wherein, when obtaining the training data set, the method further comprises:

obtaining a reference three-dimensional volume and a reference pseudo-sinogram of each of the reference objects based on the training data set; wherein the reference pseudo-sinogram is obtained by stacking collected DR images of a same reference object into a three-dimensional matrix according to a collection order, summing sinograms in each layer column by column, and normalizing a sum result;

dividing the DR images of each reference object in the training data set into a plurality of DR image groups, and each of the DR image groups comprises two DR images with a perpendicular relationship;

wherein the training to-be-trained three-dimensional volume reconstruction model using DR images with a perpendicular relationship in the training data set to obtain the preset three-dimensional volume reconstruction model comprises:

inputting the plurality of DR image groups into the to-be-trained three-dimensional volume reconstruction model, training the to-be-trained three-dimensional volume reconstruction model, and respectively obtaining a training pseudo-sinogram and a training three-dimensional volume;

calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram and the reference three-dimensional volume according to a training order;

terminating the training and obtaining the preset three-dimensional volume reconstruction model in a case that any calculated loss value is less than a target loss value, or the number of training times reaches a target number of times.

11 . The electronic device according to claim 10 , wherein the to-be-trained three-dimensional volume reconstruction model comprises an enhancement network model and a reconstruction network model;

the enhancement network model is a model based on an encoder-decoder network architecture, and configured to input a DR image group with a perpendicular relationship and output a corresponding training pseudo-sinogram; and

the reconstruction network model is an autoencoder network model for two-dimensional to three-dimensional volume mapping, and configured to input the training pseudo-sinogram and output a corresponding training three-dimensional volume.

12 . The electronic device according to claim 11 , wherein the enhancement network model to input a DR image group with a perpendicular relationship, and output a corresponding training pseudo-sinogram, comprises:

performing pattern and feature extraction on the input DR image group, and performing a cascade operation on a result of the extraction;

performing latent space feature encoding on features after the cascade operation to obtain encoded latent space features; and

decoding the encoded latent space features to obtain the training pseudo-sinogram.

13 . The electronic device according to claim 11 , wherein the reconstruction network model to input the training pseudo-sinogram and output a corresponding training three-dimensional volume, comprises:

performing feature capture and enhancement on the input training pseudo-sinogram;

encoding a result of feature enhancement, and continuously reducing the spatial resolution on the basis of keeping the number of channels unchanged; and

decoding a result of the encoding, and continuously restoring the spatial resolution to a spatial resolution of the reference three-dimensional volume to obtain the training three-dimensional volume.

14 . The electronic device according to claim 10 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

15 . The electronic device according to claim 11 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises;

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

16 . The electronic device according to claim 12 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

17 . The electronic device according to claim 13 , wherein the calculating a loss value based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume according to the training order, comprises:

calculating the loss value using a preset loss function based on the training pseudo-sinogram, the training three-dimensional volume, the reference pseudo-sinogram, and the reference three-dimensional volume,

wherein the preset loss function is a function of a weighted summation of a first loss function and a second loss function; the first loss function is configured to calculate a mean square error loss between a training pseudo-sinogram and a reference pseudo-sinogram, and the second loss function is configured to calculate a mean square error loss of a training three-dimensional volume and a reference three-dimensional volume.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (17)