Phase-cycled bSSFP PDFF mapping

MATLAB code for proton-density fat fraction (PDFF) mapping from phase-cycled balanced SSFP, with ΔB0 estimation so that fat–water swaps from field inhomogeneity can be corrected before SPARCQ reconstruction.

The method is described in:

Acikgoz BC, Mackowiak ALC, Bongiolatti-Rossi GMC, et al. Fat fraction mapping in the presence of magnetic field inhomogeneities with phase-cycled bSSFP at 3T.

Typical acquisition (as in the manuscript): TR/TE ≈ 3.4/1.7 ms, FA 35°, 3T, complex profiles [x y nPC].

Pipeline

1. Phase-correct the measured profiles (remove coil / eddy / systematic phase).
2. Fit residuals over a discretized ΔB0 range, then pick a spatially smooth field map with graph cuts.
3. Correct profiles for ΔB0 (Fourier circular shift + accumulated phase).
4. Identify very low / very high PDFF with a phase lookup table.
5. Run SPARCQ dictionary matching on the remaining voxels.

How to run

MATLAB R2023b or later. Toolboxes: Optimization (lsqnonneg, lsqlin), Image Processing (imresize, bwareaopen), Parallel Computing (parfor).

One-time: copy a working max_flow_mex (Windows: max_flow_mex.mexw64) into

third_party/matlab_bgl/private/

Do not add that private folder to the MATLAB path.

1. Put a .mat with complex profiles [x y nPC] into example data/.
2. Edit USER SETTINGS in prepare_data.m (filename, type = Liver | Knee | Calimetrix | Butters, TR, FA, …) and run it. This writes a dataset struct.
3. Set prepared_name in run_sparcq.m to that filename and run it.
4. Draw a background mask (slider + Continue).
5. Optional: in run_sparcq.m, set plot_vials = true to click phantom vials against the commercial (Calimetrix) or custom (Butters) ground-truth lists.

Prepared dataset structs and optional *_results.mat stay in example data/. Sample commercial-phantom data (shim settings and RF excitation angle) are included there. Code: https://github.com/QIS-MRI/pcbSSFP_PDFF

prepare_data.m          pack scan + algorithm parameters
run_sparcq.m            mask GUI + reconstruction + figures
lib/                    SPARCQ pipeline
third_party/hernando/   graph-cut (Hernando)
third_party/matlab_bgl/ max_flow (paste MEX in private/)
example data/           input and output .mat files

Citations

If you use this code, please cite the SPARCQ PDFF papers and, kindly, the graph-cut method and the ISMRM Fat-Water Toolbox that this reconstruction relies on (as in the manuscript, refs 10 and 28):

Graph-cut field mapping

Hernando D, Kellman P, Haldar JP, Liang ZP. Robust water/fat separation in the presence of large field inhomogeneities using a graph cut algorithm. Magn Reson Med. 2010;63(1):79-90. doi:10.1002/mrm.22177 PMID: 19859956

ISMRM Fat-Water Separation Toolbox

Hu HH, Börnert P, Hernando D, Kellman P, Ma J, Reeder S, Sirlin C. ISMRM workshop on fat-water separation: Insights, applications and progress in MRI. Magn Reson Med. 2012;68(2):378-388. doi:10.1002/mrm.24369

Toolbox: https://www.ismrm.org/workshops/FatWater12/data.htm

SPARCQ

Rossi GMC, Mackowiak ALC, Açikgöz BC, Pierzchała K, Kober T, Hilbert T, Bastiaansen JAM. SPARCQ: A new approach for fat fraction mapping using asymmetries in the phase-cycled balanced SSFP signal profile. Magn Reson Med. 2023;90(6):2348-2361. doi:10.1002/mrm.29813

Please also cite the manuscript above when using this ΔB0-corrected implementation.

License

Academic, non-commercial research only. The bundled graph-cut code is Diego Hernando / University of Wisconsin academic software; see LICENSE.