System and method for visualization and segmentation of tissue using a bayesian estimation of multicomponent relaxation values in magnetic resonance fingerprinting

1. A method for magnetic resonance fingerprinting (MRF), the method comprising:

(a) accessing with a computer system, MRF data that were acquired from a volume in a subject using an MRI system;

(b) accessing with the computer system, a dictionary of signal evolutions;

(c) selecting with the computer system, a plurality of regions-of-interest (ROIs) in the MRF data, wherein the ROIs correspond to different tissue types;

(d) generating a first series of tissue parameter estimates by inputting the dictionary and the MRF data contained in the plurality of ROIs to a multicomponent Bayesian framework, generating output as tissue parameter estimates for each ROI;

(e) computing a plurality of probability distributions from the first series of tissue parameter estimates, wherein the plurality of probability distributions comprises a probability distribution for each different tissue type;

(f) creating a reduced dictionary by removing entries from the dictionary having tissue parameter values not contained within the computed probability distributions;

(g) generating a second series of tissue parameter estimates by inputting the reduced dictionary and the MRF data to a multicomponent Bayesian framework, generating output as tissue parameter estimates for each voxel in the volume; and

(h) generating a tissue probability map for each different tissue type from the second series of tissue parameter estimates.

2. The method of claim 1 , wherein the tissue probability map for a given tissue type is generated by computing for each voxel a weighted sum of tissue parameters in the second series of tissue parameter estimates and multiplying the weighted sum by the probability distribution for the given tissue type.

3. The method as recited in claim 1 , wherein computing the probability distribution for each different tissue type comprises inputting the first series of tissue parameter estimates to a Gaussian mixture model, generating output as the plurality of probability distributions.

4. The method as recited in claim 3 , further comprising selecting each probability distribution for each different tissue type from the plurality of probability distributions based on a maximum mixing probability of each of the plurality of probability distributions.

5. The method as recited in claim 1 , wherein step (g) includes storing a maximum probability for each dictionary entry in the reduced dictionary, wherein the maximum probability indicates the probability distribution to which a given dictionary entry is most likely to belong.

6. The method as recited in claim 1 , wherein the ROIs in the plurality of ROIs each consist of a single voxel.

7. The method as recited in claim 6 , wherein each ROI corresponds to a different single tissue type.

8. The method as recited in claim 6 , wherein at least one ROI corresponds to a sub-voxel mixture of two or more tissue types.

9. The method as recited in claim 1 , wherein the ROIs in the plurality of ROIs each comprise one or more voxels.

10. The method as recited in claim 9 , wherein each ROI corresponds to a different single tissue type.

11. The method as recited in claim 9 , wherein at least one ROI corresponds to a sub-voxel mixture of two or more tissue types.

12. The method of claim 1 , wherein the different tissue types comprise white matter, gray matter, and cerebrospinal fluid.

13. The method as recited in claim 12 , wherein at least one member of the series of variable sequence blocks differs from at least one other member of the series of variable sequence blocks in at least two sequence block parameters.

14. The method of claim 1 , wherein the MRF data were acquired with the MRI system in a series of variable sequence blocks to cause one or more resonant species in the subject to simultaneously produce individual magnetic resonance signals.