Preclinical magnetic resonance imaging of proinflammatory epicardial adipose tissue: accelerated methods for simultaneous fatty acid composition and relaxation parameter mapping with relationships to tissue biomarkers

Figures

• 

  Fig. 1. Image reconstruction framework. The HD-PROST-based reconstruction alternates between two subproblems to recover denoised multicontrast images from undersampled radial k-space data. Optimization 2: The initial image estimate X0, generated via NUFFT, is divided into overlapping patches. Patches from the same spatial location across all contrast dimensions are unfolded into 3D tensors (T~p) and denoised using HOSVD followed by adaptive hard thresholding. Denoised patches are reassembled into a tensor T. Optimization 1: A joint reconstruction update enforces data consistency with the measured k-space data Y and regularization toward the denoised image T, yielding an updated image X. The Lagrange multiplier (b) is updated after each iteration. This process is repeated for NADMM iterations to produce the final reconstructed image series XFinal. σopt optimal hard threshold. HOSVD higher-order singular value decomposition, NUFFT nonuniform fast Fourier transform, 3D three dimensional, HD-PROST high-dimensionality undersampled patch-based reconstruction

• 

  Fig. 2. Pulse sequence diagram. (A) Inversion recovery interleaved multi-echo gradient-echo pulse sequence for M TIs and N TEs, showing the acquisition of a single radial spoke at TEn and TEn+10 across multiple TIs, followed by RR pauses. (B) Example k-space trajectories with golden-angle rotation in-plane and through contrast dimensions, illustrating six spokes acquired at different TEs and TIs. Ns number of spokes, NTE number of total echo times, TI inversion times, TE echo times

• 

  Fig. 3. Error analysis of retrospectively undersampled images. MAE between (A) SFA, MUFA, and PUFA fractions, (B) T1, (C) R2*, and (D) PDFF values computed from retrospectively undersampled phantom images and fully sampled acquisitions. (E) SSIM index between denoised retrospectively undersampled and fully sampled images. FAC fatty acid composition, PDFF proton density fat fraction, SFA saturated fatty acid, MUFA monounsaturated fatty acid, PUFA polyunsaturated fatty acid, MAE mean absolute error, SSIM structural similarity

• 

  Fig. 4. Phantom layout and parameter maps. (A) Layout of oil and gadolinium phantoms used for FAC and T1 validation, respectively. The FAC phantom consists of five 1 mL oil mixtures submerged in 5 mL of water, while the T₁ phantom contains six 1 mL vials of water with Gd concentrations ranging from 0.1 to 0.5 mM. (B) Parameter maps of PDFF, SFA/MUFA/PUFA fractions, R2*, and T1 for fully sampled and rate-9.6 accelerated acquisitions, overlaid on the phantom images. (C) Linear regression analysis between joint FAC and T1 measured values and reference values for fully sampled and rate-9.6 accelerated acquisitions. Pearson r values are reported. Gd gadolinium, FAC fatty acid composition, PDFF proton density fat fraction, SFA saturated fatty acid, MUFA monounsaturated fatty acid, PUFA polyunsaturated fatty acid

• 

  Fig. 5. Comparison of undersampled and HD-PROST reconstructed in vivo images. Nonuniform fast Fourier transform undersampled and HD-PROST reconstructed images acquired at acceleration rates 3.7 and 9.6 for (A) TE1, TI1, (B) TE1, TI15, and (C) TE4, TI30 in the same mouse with high-fat high-sucrose diet-induced obesity. HD-PROST high-dimensionality undersampled patch-based reconstruction, TI inversion times, TE echo times

• 

  Fig. 6. In vivo parametric maps from a representative HFHSD mouse. Images from a mid-ventricular axial slice showing: (A) T1-weighted reference image, (B) water (W) and fat (F) signal images (yellow arrows = EAT, green arrows = SAT), (C) maps of off-resonance (ψ), R2*, and T1, and (D) fat parameter maps including PDFF, SFA fraction, MUFA fraction, and PUFA fraction. Parameter maps are overlaid on manually contoured EAT and SAT regions. HFHSD, high-fat high-sucrose diet. EAT epicardial adipose tissue, SAT subcutaneous adipose tissue, PDFF proton density fat fraction, SFA saturated fatty acid, MUFA monounsaturated fatty acid, PUFA polyunsaturated fatty acid

• 

  Fig. 7. Diet and SGLT2 inhibitor-induced differences in adipose tissue MRI parameters. Tukey boxplots of absolute SFA fraction, MUFA fraction, PUFA fraction, PDFF, T1, and R2* values for (A) EAT and (B) SAT in mice fed an HFHSD (n = 20), HFHSD+EMPA (n = 17), or HFD (n = 16) for 18 weeks. (C) EAT index values for SFA, MUFA, PUFA, and PDFF in the same groups. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001. HFHSD, high-fat high-sucrose diet; EMPA, empagliflozin; HFD, high-fat diet. EAT epicardial adipose tissue, SAT subcutaneous adipose tissue, PDFF proton density fat fraction, SFA saturated fatty acid, MUFA monounsaturated fatty acid, PUFA polyunsaturated fatty acid, SGLT2 sodium-glucose cotransporter-2 inhibitor

• 

  Fig. 8. Epicardial adipose tissue histology. (A) F4/80 (green) and DAPI (blue) staining show increased macrophage infiltration in HFD and HFHSD mice, which is reduced with EMPA treatment (left column). Merged F4/80, DAPI, and WGA (white) images highlight crown-like structures of macrophages surrounding adipocytes (red arrows) in HFD and HFHSD conditions (middle column). Scale bar: 20 µm. (B) Hematoxylin and eosin (H&E) staining reveals larger adipocytes in HFD and HFHSD conditions, while EMPA-treatment displays smaller, more uniform adipocytes. Images in (A) and (B) were obtained from different mice within the same experimental groups. Scale bar: 100 µm. WGA, wheat germ agglutinin. HFHSD high-fat high-sucrose diet, EMPA empagliflozin, HFD high-fat diet

• 

  Fig. 9. Relationships between EAT MRI parameters and tissue inflammation markers. Correlations between EAT MRI parameters – (A) SFA index, (B) MUFA index, (C) PUFA index, (D) PDFF index, (E) R2*, and (F) T1 – and tissue measures, including macrophage density (Mφ/adipocyte), adipocyte size (µm²), and cytokine levels (pg/µg total protein). Spearman’s r and p-values are reported for each plot. HFHSD, high-fat high-sucrose diet. EAT epicardial adipose tissue, SAT subcutaneous adipose tissue, FAC fatty acid composition, PDFF proton density fat fraction, SFA saturated fatty acid, MUFA monounsaturated fatty acid, PUFA polyunsaturated fatty acid