Method and system for denoising CT images using a neural network
View Patent ↗First and second substantially independent identically distributed half scans are obtained; the first substantially independent identically distributed half scan is used as training data to train a machine learning-based system, and the second substantially independent identically distributed half scan is used as label data to train a machine learning-based system. This produces a trained machine learning-based system.
1. A method of generating an image denoising system, the method comprising:
acquiring single scan data obtained from a single scan of a subject, the single scan data being count-domain projection data;
identically distribute the acquired single scan data to generate first and second substantially independent partial scan data, wherein the first partial scan data is generated by applying a thinning model to the projection data and the second partial scan data is generated by subtracting the generated first partial scan data from the projection data; and
training a machine learning-based system based on the generated first substantially independent, identically distributed, partial scan data as input training data, and the generated second substantially independent, identically distributed, partial scan data as label data to produce a trained machine learning-based system.
2. The method as claimed in claim 1 , wherein each of the first and second partial scan data corresponds to a half-dose scan.
3. The method as claimed in claim 1 , wherein the machine learning-based system comprises a neural network and the trained machine learning-based system comprises a trained neural network.
4. The method as claimed in claim 3 , wherein the neural network comprises a deep neural network.
5. The method as claimed in claim 1 , wherein the acquired single scan data includes court-domain projection data.
6. The method as claimed in claim 1 , wherein the acquired single scan data includes image-domain reconstructed images.
7. The method as claimed in claim 1 , further comprising adding Gaussian noise to at least one of the generated first and second substantially independent, identically distributed, partial scan data prior to the training step.
8. The method of claim 1 , wherein the generated first and second substantially independent, identically distributed, partial scan data are first and second nearly independent, identically distributed, partial scan data.
9. The method of claim 1 , wherein the generated first and second substantially independent, identically distributed, partial scan data are first and second virtually independent, identically distributed, partial scan data.
10. The method of claim 1 , wherein the generated first and second substantially independent identically distributed partial scan data are first and second completely independent, identically distributed, partial scan data.
11. A trained machine learning-based system produced according to the method of claim 1 .
12. A system for generating an image demising system, comprising:
processing circuitry configured to
acquire single scan data obtained from a single scan of a subject, the single scan data being count-domain projection data;
identically distribute the acquired single scan data to generate first and second substantially independent partial scan data, wherein the first partial scan data is generated by applying a thinning model to the projection data and the second partial scan data is generated by subtracting the generated first partial scan data from the projection data; and
train a machine learning-based system based on the generated first substantially independent, identically distributed, partial scan data as input training data, and the generated second substantially independent, identically distributed, partial scan data as label data to produce a trained machine learning-based system.
13. The system as claimed in claim 12 , wherein each of the first and second partial scan data corresponds to a half-dose scan.
14. The system as claimed in claim 12 , wherein the machine learning-based system comprises a neural network and the trained machine learning-based system comprises a. trained neural network.
15. The system as claimed in claim 14 , wherein the neural network comprises a deep neural network.
16. The system as claimed in claim 12 , wherein the acquired single scan data includes count-domain projection data.
17. The system as claimed in claim 12 , wherein the acquired single scan data includes image-domain reconstructed images.
18. The system as claimed in claim 12 , wherein the processing circuitry is further configured to add Gaussian noise to at least one of the generated first and second substantially independent, identically distributed, partial scan data prior to the processing circuitry training the machine learning-based system.
19. The system as claimed in claim 12 , wherein the generated first and second substantially independent, identically distributed, partial scan data are first and second completely independent, identically distributed, partial scan data.
20. A trained machine learning-based system produced using the system according to claim 12 .
21. A medical image processing apparatus, comprising:
processing circuitry configured to
receive medical image data; and
perform denoising processing by applying the medical image data to a trained machine learning-based system, wherein the trained machine learning-based system is generated by (a) acquiring single scan data obtained from a single scan of a subject, the single scan data being count-domain projection data, (b) identically distributing the acquired single scan data to generate first and second substantially independent partial scan data, wherein the first partial scan data is generated by applying a thinning model to the projection data and the second partial scan data is generated by subtracting the generated first partial scan data from the projection data; and (c) training a machine learning-based system based on the generated first substantially independent, identically distributed, partial scan data as input training data and the generated second substantially independent, identically distributed, partial scan data as label data to produce the trained machine learning-based system.
22. A medical image processing method, comprising:
receiving medical image data; and
performing denoising processing by applying the medical image data to a trained machine learning-based system, wherein the trained machine learning-based system is generated by (a) acquiring single scan data obtained from a single scan of a subject, the single scan data being count-domain projection data, (b) identically distributing the acquired single scan data to generate first and second substantially independent partial scan data, wherein the first partial scan data is generated by applying a thinning model to the projection data and the second partial scan data is generated by subtracting the generated first partial scan data from the projection data; and (c) training a machine learning-based system based on the generated first substantially independent, identically distributed, partial scan data as input training data and the generated second substantially independent, identically distributed, partial scan data as label data to produce the trained machine learning-based system.