Fast group matching for magnetic resonance fingerprinting reconstruction

Patent metadata

Field	Value
Publication number	US-9964616-B2
Application number	US-201514711815-A
Country	US
Kind code	B2
Filing date	May 14, 2015
Priority date	May 30, 2014
Publication date	May 8, 2018
Grant date	May 8, 2018

Abstract

Methods, apparatus, and other embodiments associated with producing a quantitative parameter map using magnetic resonance fingerprinting (MRF) are described. One example apparatus includes a data store that stores a grouped set of MRF signal evolutions, including a group representative signal and a low-rank representative, a set of logics that collects a received signal evolution from a tissue experiencing nuclear magnetic resonance (NMR) in response to an MRF excitation, a correlation logic that computes a correlation between a portion of the received signal evolution and a portion of a group representative signal, a pruning logic that generates a pruned grouped set, and a matching logic that determines matching quantitative parameters based on the received signal evolution and the low-rank representative.

First claim

What is claimed is:

1. A non-transitory computer-readable storage medium storing computer executable instructions that when executed by a computer control the computer to perform a method for producing a quantitative parameter map, comprising:
accessing a comprehensive dictionary of predicted magnetic resonance fingerprinting (MRF) signal evolutions, where a signal evolution includes complex values with an arbitrary phase relationship;
generating a grouped dictionary based on the comprehensive dictionary, where the grouped dictionary includes a plurality of groups, where a group includes a plurality of correlated signal evolutions, a group representative signal that represents the group, and a group low-rank representative;
accessing patient data, where the patient data includes a patient MRF signal evolution, where the patient MRF signal evolution includes complex values with an arbitrary phase relationship;
comparing the patient MRF signal evolution with the group representative signal from a threshold number of the plurality of groups;
upon determining that the group representative signal from the threshold number does not match the patient MRF signal evolution to within a desired threshold,
generating a pruned dictionary by removing from consideration, from the grouped dictionary, the groups represented by the threshold number of representative signals that do not match the patient MRF signal evolution;

selecting a matched signal evolution by matching, to within a threshold quality of fit, the patient MRF signal evolution with a signal evolution in the pruned dictionary, where the matching is based on the group low-rank representative, where the threshold quality of fit is a dynamic, adaptive threshold;
determining a quantitative parameter for the patient MRF signal evolution based on the matched signal evolution; and
generating a patient image based, at least part, on the quantitative parameter.

2. The non-transitory computer-readable storage medium of claim 1 , where generating the grouped dictionary comprises:
selecting a first MRF signal evolution from the comprehensive dictionary;
defining a group based on the first MRF signal evolution, where the group includes the first MRF signal evolution;
comparing the first MRF signal evolution with a second, different MRF signal evolution selected from the comprehensive dictionary;
upon determining that the second MRF signal evolution is within a threshold correlation of the first MRF signal evolution,
assigning the second MRF signal evolution to the group;

upon determining that the second MRF signal evolution is not within the threshold correlation of the first MRF signal evolution,
selecting a different MRF signal evolution from the comprehensive dictionary to compare with the first signal;

generating a group representative signal for the group; and
computing a group low-rank representative for the group, based, at least in part, on a PCA approach using singular value decomposition.

3. The non-transitory computer-readable storage medium of claim 1 , the method comprising identifying one or more MR parameters associated with a voxel in the patient image that produced the patient MRF signal evolution based, at least in part, on the quantitative parameter.

4. The non-transitory computer-readable storage medium of claim 3 , where the one or more MR parameters include longitudinal relaxation time T1, transverse relaxation time T2, main magnetic field map B0, or proton density ρ.

5. The non-transitory computer-readable storage medium of claim 2 , where determining that the second MRF signal evolution is within a threshold correlation of the first MRF signal evolution is a function of sparse methods including K-way partitioning, or greedy choice grouping.

6. The non-transitory computer-readable storage medium of claim 2 , where generating the group representative signal for the group includes calculating the mean signal across the plurality of correlated signal evolutions in the group, or calculating a statistics-based estimation from the plurality of correlated signal evolutions in the group.

7. The non-transitory computer-readable storage medium of claim 2 , where the grouped dictionary includes all the elements of the comprehensive dictionary, or a subset of the elements of the comprehensive dictionary, where the size of the subset is based, at least in part, on a stopping criterion.

8. The non-transitory computer-readable storage medium of claim 2 , where selecting the first MRF signal evolution from the comprehensive dictionary includes randomly selecting an MRF signal evolution from the comprehensive dictionary, or selecting a mean signal evolution for a tissue type.

9. The non-transitory computer-readable storage medium of claim 2 , where the grouped dictionary includes a number M of MRF signal evolutions, where M is an integer greater than 1, and where the M MRF signal evolutions are evenly spread across a number N groups, where N is an integer greater than 0.

10. The non-transitory computer-readable storage medium of claim 2 , where the grouped dictionary includes a number M of MRF signal evolutions, where M is an integer greater than 1, and where the M MRF signal evolutions are spread unevenly across a number N groups, where N is an integer greater than 0.

11. The non-transitory computer-readable storage medium of claim 1 , where generating the pruned dictionary is a function of a pruning criterion, where the pruning criterion is determined through a relative correlation threshold or an absolute correlation threshold.

12. The non-transitory computer-readable storage medium of claim 11 , where the pruning criterion is a dynamic, adaptive pruning criterion based, at least in part, on a dynamic best option correlation between the patient MRF signal evolution and a group representative signal evolution, the level of redundancy across the group representative signals in the grouped dictionary, the level of compression within groups in the pruned dictionary, or the level of compression within groups in the grouped dictionary.

13. The non-transitory computer-readable storage medium of claim 12 , where the pruning criteria is a relative pruning criteria of 5×10 −3 below a best group match for the patient MRF signal evolution.

14. A method for generating a pruned dictionary for matching signals acquired using magnetic resonance fingerprinting (MRF), comprising:
accessing a grouped MRF signal dictionary that stores signals characterized by MRF dictionary signal equations, where a group in the grouped MRF signal dictionary includes a plurality of MRF signal evolutions having a threshold correlation, a representative signal evolution, and a low-rank representative;
acquiring a patient MRF signal evolution from an MRF apparatus that repetitively and variably samples a space associated with human tissue to produce the MRF signal;
computing a correlation between the patient MRF signal evolution and the representative signal evolution of a given group; and
upon determining that the correlation between patient MRF signal evolution and the representative signal evolution is not within a threshold correlation:
generating a pruned group dictionary by removing from consideration the group associated with the representative signal evolution from the grouped MRF signal dictionary; and
generating an image using the MRF signal evolutions and correlation information determined using the pruned group dictionary.

15. The method of claim 14 , where generating the pruned group dictionary comprises:
computing a correlation between the patient MRF signal evolution and

Assignees

• Univ Case Western Reserve
• Massachusetts Gen Hospital

Inventors

• Cauley Stephen
• Griswold Mark
• Setsompop Kawin
• Wald Lawrence

Classifications

• G01R33/4828

  Resolving the MR signals of different chemical species, e.g. water-fat imaging · CPC title

• G01R33/50 Primary

  based on the determination of relaxation times {, e.g. T1 measurement by IR sequences; T2 measurement by multiple-echo sequences} · CPC title