Trajectory correction enables free-running chemical shift encoded imaging for accurate cardiac proton-density fat fraction quantification at 3T

Fig. 1. Comparison of PDFF (A) and R2* (B) bias and absolute error between 8 echoes in monopolar compared to 13 echoes in bipolar over synthetic CSE-CMR volumes with SNR = 50 and 10. Mean and standard deviation PDFF and R2* bias and absolute error were averaged along the B0 off-resonance and repetition axes. CSE chemical shift encoded, CMR cardiovascular magnetic resonance , PDFF proton-density fat fraction, SNR signal-to-noise ratio.

Fig. 2. Trajectory correction using system-specific GIRF: (A) measurement of the GIRF and the magnitude of its Fourier transform (GFRF); (B) k-space trajectory correction pipeline (C) applied for the first four echoes in bipolar mode. (D) Converting distorted gradients into k-space trajectory showed up a mean shift superior to 1 for even echoes. GIRF-corrected trajectory prevents this error at the image reconstruction level. GIRF gradient impulse response function, GFRF gradient frequency response function.

Fig. 3. Free-running CSE-CMR echo times repartition depending on the multi-echo scheme, before and after GIRF correction. Monopolar echo spacing was longer than the fat-water in-phase/out-of-phase delay at 3T. In bipolar mode without correction, strong artifacts appeared on even echo images but after GIRF correction, image quality was restored. CSE chemical shift encoded, GIRF gradient impulse response function, CMR cardiovascular magnetic resonance.

Fig. 4. PDFF (A) and B0 off-resonance (B) measurements from in vitro experiments with monopolar and bipolar with or without GIRF correction. GIRF gradient impulse response function, PDFF proton-density fat fraction.

Fig. 5. Impact of GIRF correction on free-running bipolar CSE-CMR on in-vivo data. Artifacted even echoes (second echo presented only) are highlighted in red resulting in severe swaps between fat and water reconstructed volumes, as pointed by the red arrows. Radial artifacts were reduced in GIRF-corrected images, corresponding to an almost in-phase image in vivo. CSE chemical shift encoded, GIRF gradient impulse response function, CMR cardiovascular magnetic resonance.

Fig. 6. Quantitative PDFF and R2*, B0 off-resonance maps in axial view resulting from monopolar echoes or bipolar ones with/without GIRF correction revealed the propagation of image artifacts within the fat-water separation process. Across the different phases of the cardiac, inadequate monopolar echo spacing as well as uncorrected bipolar data led to spurious PDFF overestimation with fat-water swaps. GIRF gradient impulse response function, PDFF proton-density fat fraction.

Fig. 7. Quantitative PDFF and R2*, B0 off-resonance maps in coronal view resulting from bipolar with or without GIRF correction and monopolar echoes revealed the propagation of image artifacts within the fat-water separation process. Across the different phases of the respiratory cycle, inadequate monopolar echo spacing as well as uncorrected bipolar data led to spurious PDFF overestimation. The dotted white line represents the diaphragm level at expiration state. GIRF gradient impulse response function, PDFF proton-density fat fraction.

Fig. 8. Comparison of PDFF measurements over different regions of interest across the volunteers between (A) bipolar echoes with or without GIRF correction, (B) monopolar or bipolar GIRF corrected echoes. Boxplots showed the consistency of PDFF measurements along cardiac and respiratory cycles. GIRF gradient impulse response function, LV: left ventricle, PDFF proton-density fat fraction.

Fig. 9. (A) PDFF in epicardial, paracardial, and subcutaneous fat across healthy volunteers and type-2 diabetic patients. (B) Comparison of PDFF maps between two patients (P01 and P03) and two healthy volunteers (V05 and V09) that presented both types of fat. PDFF maps are overlaid on anatomical images in four-chamber views for each corresponding subject. Arrows highlight each cardiac fat depot, green for EAT and white for PAT. EAT epicardial adipose tissue, PAT paracardial adipose tissue, PDFF proton-density fat fraction, SAT subcutaneous adipose tissue.