Image data restoration apparatus and image data restoration method

1. An image data restoration apparatus, comprising:

processing circuitry configured to:

acquire, with respect to a target, a plurality of items of observation data respectively corresponding to a plurality of channels used in a computation;

select a first channel subset and a second channel subset, the first channel subset including one or more channels of the plurality of channels, the second channel subset including, of the plurality of channels, channels larger in number than the first channel subset; and

iteratively execute a computation to optimize a fitting score between first observation data corresponding to the first channel subset and first intermediate image data based on the first observation data, and execute, after the iterative computation satisfies a predetermined condition, a computation to optimize a fitting score between second observation data corresponding to the second channel subset and second intermediate image data based on the second observation data to generate restored image data and output the restored image data.

2. The image data restoration apparatus according to claim 1 , wherein

the second channel subset is the plurality of channels, and

the second observation data is the plurality of items of observation data.

3. The image data restoration apparatus according to claim 1 , wherein

the fitting score includes a data fitting score and a prior knowledge fitting score, the data fitting score indicating a fitting score of the restored image data based on a restoration model with respect to the observation data, the prior knowledge fitting score indicating a fitting score of the restored image data with respect to a prior knowledge model of the restored image data,

the processing circuitry is further configured to:

execute a computation to optimize the data fitting score and the prior knowledge fitting score.

4. The image data restoration apparatus according to claim 1 , further comprising:

a memory which stores a restoration model that associates the acquired observation data with the restored image data, and a prior knowledge model that indicates prior knowledge about the restored image data.

5. The image data restoration apparatus according to claim 1 , wherein

the processing circuitry is further configured to:

determine whether the iterative computation satisfies a predetermined condition.

6. The image data restoration apparatus according to claim 5 , wherein the predetermined condition is that an evaluation function for the first intermediate image data exceeds a threshold value.

7. The image data restoration apparatus according to claim 5 , wherein the predetermined condition is that an iteration count of the iterative computation reaches a predetermined upper limit value.

8. The image data restoration apparatus according to claim 3 , wherein

the processing circuitry is further configured to:

scale the data fitting score in accordance with the number of the selected channels and the number of the plurality of channels corresponding to the acquired observation data.

9. The image data restoration apparatus according to claim 3 , wherein

the processing circuitry is further configured to:

scale the data fitting score in accordance with a ratio between a norm of observation data corresponding to the selected channels and a norm of the acquired observation data.

10. The image data restoration apparatus according to claim 1 , wherein

the processing circuitry is further configured to:

convert the plurality of channels corresponding to the acquired observation data into a plurality of virtual channels by applying an orthogonal transformation, and

select a virtual channel corresponding to the observation data used in the computation, based on eigenvalues of the observation data corresponding to the plurality of virtual channels.

11. The image data restoration apparatus according to claim 1 , wherein

the computation includes a first computation to optimize the restored image data, and a second computation to optimize internal data, and

the processing circuitry is further configured to:

iteratively execute the computation by alternately executing the first computation and the second computation.

12. The image data restoration apparatus according to claim 3 , wherein

the plurality of channels correspond to observation data observed for respective coils in an MRI apparatus, and

the data fitting score includes a term that evaluates an L2-norm or a Frobenius norm of a difference between the observation data and data obtained by applying the restoration model to the restored image data.

13. The image data restoration apparatus according to claim 3 , wherein

the plurality of channels correspond to observation data observed for respective spectra in a multispectral camera or a spectral X-ray apparatus, and

the data fitting score includes a term that evaluates an L2-norm of a difference between the observation data and data obtained by applying the restoration model to the restored image data.

14. The image data restoration apparatus according to claim 3 , wherein

the processing circuitry is further configured to:

obtain time-series observation data observed with respect to the target across multiple times, and

the prior knowledge fitting score includes a term that evaluates an L1-norm of a difference between the restored image data at a first time and the restored image data at a second time different from the first time.

15. An image data restoration method, comprising:

acquiring, with respect to a target, a plurality of items of observation data respectively corresponding to a plurality of channels used in a computation;

selecting a first channel subset and a second channel subset, the first channel subset including one or more channels of the plurality of channels, the second channel subset including, of the plurality of channels, channels larger in number than the first channel subset; and

iteratively executing a computation to optimize a fitting score between first observation data corresponding to the first channel subset and first intermediate image data based on the first observation data, and executing, after the iterative computation satisfies a predetermined condition, a computation to optimize a fitting score between second observation data corresponding to the second channel subset and second intermediate image data based on the second observation data to generate restored image data and outputting the restored image data.