Data-driven gating based upon grouping features in short image data segments

Patent metadata

Field	Value
Publication number	US-12548220-B2
Application number	US-202217965289-A
Country	US
Kind code	B2
Filing date	Oct 13, 2022
Priority date	Apr 27, 2022
Publication date	Feb 10, 2026
Grant date	Feb 10, 2026

Abstract

A method, apparatus, and computer instructions stored on a computer-readable medium perform latent image feature extraction by performing the functions of receiving image data acquired during an imaging of a patient, wherein the image data includes motion by the patient during the imaging; segmenting the image data to include M image data segments corresponding to at least N motion phases having shorter durations than a duration of the motion by the patient during the imaging, wherein M is a positive integer greater than or equal to a positive integer N; producing, from the M image data segments, at least N latent feature vectors corresponding to the motion by the patient during the imaging; and performing a gated reconstruction of the N motion phases by reconstructing the image data based on the at least N latent feature vectors.

First claim

The invention claimed is:

1 . A medical imaging method for data-driven reconstruction, comprising:
receiving sinogram image data acquired during an imaging of a patient, wherein the sinogram image data includes motion by the patient during the imaging; segmenting the sinogram image data into M image data segments each having a shorter duration than a duration of N motion phases of the motion by the patient during the imaging, wherein M is a positive integer greater than or equal to N, which is a positive integer; producing, from the M sinogram image data segments, N sets of latent feature vectors corresponding to the N motion phases of the motion by the patient during the imaging, wherein each of the latent feature vectors represents a feature of one of the M sinogram image data segments; and performing a reconstruction of the N motion phases by reconstructing, on a set-by-set basis, the sinogram image data associated with the N sets of latent feature vectors.

2 . The method according to claim 1 , wherein the producing, further comprises clustering the M sinogram image data segments into N sets of latent feature vectors corresponding to the N motion phases of the motion by the patient during the imaging; and
wherein performing the reconstruction of the N motion phases further comprises performing the reconstruction of the N motion phases by reconstructing the image data based on the clustering of the M sinogram image data segments into N sets of latent feature vectors corresponding to the N motion phases of the motion by the patient during the imaging.

3 . The method according to claim 2 , wherein the clustering is performed by using at least one of a Gaussian Mixture Model, Spectral Clustering, and a support vector machine method.

4 . The method according to claim 2 , where the clustering is performed by using at least one of a logistic regression, a Naive Bayes method, and a decision-tree method.

5 . The method according to claim 1 , wherein durations of first and second motion phases of the N motion phases are different.

6 . The method according to claim 1 , wherein the producing further comprises training an untrained neural network to produce, from the M sinogram image data segments, the N sets of latent feature vectors.

7 . The method according to claim 6 , wherein the training comprises training the untrained neural network, which is at least one of an autoencoder and a variational autoencoder.

8 . The method according to claim 6 , wherein training the untrained neural network to produce, from the M sinogram image data segments, the N sets of latent feature vectors further comprises training the untrained neural network using a loss function.

9 . The method according to claim 8 , wherein the loss function is based on a metric including at least one of a mean-squared-error (MSE), a mean-absolute-error (MAE), and a root-mean-square error (RMSE).

10 . The method according to claim 6 , wherein the untrained neural network is trained by using only data specific to a current patient.

11 . The method according to claim 1 , wherein each of the N motion phases are in a range of 0.05 to 2.0 seconds.

12 . The method according to claim 1 , wherein N is at least four and each of the N motion phases corresponds to a different phase of breathing.

13 . The method according to claim 1 , wherein each of the N motion phases corresponds to a different cardiac phase.

14 . The method according to claim 1 , wherein the segmenting the sinogram image data into M image data segments further comprises extracting list mode data into the M image data segments.

15 . The method according to claim 1 , wherein the producing further comprises producing the N sets of latent feature vectors from respective differences between the M sinogram image data segments and M subsequent image data segments.

16 . The method according to claim 1 , wherein the reconstruction is a reconstruction without scatter correction.

17 . The method according to claim 1 , wherein the reconstruction is a reconstruction with scatter correction.

18 . The method according to claim 1 , wherein the M sinogram image data segments correspond to a single patient.

19 . The method according to claim 1 , wherein N is at least two and a first of the N motion phases corresponds to an end of expiration and a second of the N motion phases corresponds to an end of inspiration.

20 . The method according to claim 1 , wherein N is at least one and corresponds to a quiescent cardiac phase.

21 . An image processing apparatus, comprising:
processing circuitry configured to perform the method of claim 1 .

22 . A non-transitory computer-readable medium having instructions stored therein that, when executed by at least one processor, cause the at least one processor to perform the method of claim 1 .

23 . A medical imaging method for data-driven gating, comprising:
receiving first image data acquired during a first imaging of a first patient, wherein the first image data includes motion by the first patient during the first imaging; segmenting the first image data into M image data segments each having a shorter duration than a duration of N motion phases of the motion by the first patient during the first imaging, wherein M is a positive integer greater than or equal to N, which is a positive integer; producing, from the M image data segments, a trained neural network for generating latent feature vectors, wherein each of the latent feature vectors represents a feature of one of the M image data segments and corresponds to the motion by the first patient during the first imaging; receiving second image data acquired during a second imaging of a second patient, wherein the second image data includes motion by the second patient during the second imaging; segmenting the second image data into second-patient image data segments; inputting the second-patient image data segments to the trained neural network to produce second-patient latent feature vectors, wherein each of the second-patient latent feature vectors represents a feature of one of the second-patient image data segments and corresponds to the motion by the second patient during the second imaging; and performing a reconstruction of the N motion phases by reconstructing, on a set-by-set basis, the second image data based on the second-patient latent feature vectors.

24 . The method according to claim 23 , wherein the first and second image data comprises sinogram image data.

25 . The method according to claim 23 , wherein the first and second image data comprises image domain image data.

26 . A medical imaging method for data-driven reconstruction, comprising:
receiving image data acquired during an imaging of a patient, wherein the image data includes motion by the patient during the imaging; segmenting the image data into M image data segments having a shorter duration than a duration of N motion phases of the motion by the patient during the imaging, wherein M is a positive integer greater than or equal to N which is a positive integer; producing, from the M image data segments, N sets of latent feature vectors corresponding to the N motion phases of the motion by the patient during the imaging by training an untrained variational autoencoder to produce latent feature vectors from the M image data segments, wherein each of the latent feature vectors repr

Assignees

• Univ California
• Canon Medical Systems Corp

Inventors

• Qi Jinyi
• Li Tiantian
• Xie Zhaoheng
• Qi Wenyuan
• Yang Li
• Chan Chung
• Asma Evren

Classifications

• G06V10/82

  using neural networks · CPC title

• G06T2207/10104

  Positron emission tomography [PET] · CPC title

• G06T7/20

  Analysis of motion (motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51) · CPC title

• G06V20/49

  Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes · CPC title

• G06T2207/20081

  Training; Learning · CPC title