Abstract
This work introduced a stack-of-radial multi-echo asymmetric-echo MRI sequence for free-breathing liver volumetric acquisition. Regularized model-based reconstruction was implemented in Berkeley Advanced Reconstruction Toolbox (BART) to jointly estimate all physical parameter maps (water, fat, R2, and B0 field inhomogeneity maps) and coil sensitivity maps from self-gated k-space data. Specifically, locally low rank and temporal total variation regularization were employed directly on physical parameter maps. The proposed free-breathing radial technique was tested on a water/fat & iron phantom, a young volunteer, and obesity/diabetes/hepatic steatosis patients. Quantitative fat fraction and R2 accuracy were confirmed by comparing our technique with the reference breath-hold Cartesian scan. The multi-echo radial sampling sequence achieves fast k-space coverage and is robust to motion. Moreover, the proposed motion-resolved model-based reconstruction allows for free-breathing liver fat and R2* quantification in multiple motion states. Overall, our proposed technique offers a convenient tool for non-invasive liver assessment with no breath holding requirement.
Introduction
Quantifying liver fat (Proton Density Fat Fraction, or PDFF) and iron concentration ($R_2^*$) is critical for diagnosing conditions like non-alcoholic fatty liver disease (NAFLD) and hemochromatosis. The current clinical standard is a multi-echo "Dixon" MRI scan, which typically requires patients to hold their breath for 15–20 seconds. This requirement is often prohibitive for patients with obesity, diabetes, or severe illness. Motion during these scans leads to artifacts that compromise quantitative accuracy. Furthermore, traditional Cartesian scans are sensitive to $B_0$ field inhomogeneities, which can cause errors in water-fat separation.
Methods
The proposed solution, termed MERLOT, consists of a novel acquisition sequence and a sophisticated reconstruction pipeline.
Acquisition: Multi-Echo Radial FLASH
Instead of traditional grid-based (Cartesian) sampling, the method utilizes a "stack-of-radial" approach. Data is sampled along radial spokes at different angles using a Golden Angle increment. This sampling pattern is inherently robust to motion, as motion artifacts manifest as manageable streaks rather than coherent "ghosting." Asymmetric echoes are employed to minimize the repetition time (TR), enabling faster volumetric coverage.
Reconstruction: Regularized Model-based Joint Estimation
The reconstruction is formulated as a nonlinear inverse problem. Rather than reconstructing individual images and then fitting a model, the algorithm jointly estimates maps for water, fat, $R_2^*$, $B_0$ field, and coil sensitivities directly from the raw $k$-space data.
- Regularization: To handle scanning noise, Locally Low Rank (LLR) and Temporal Total Variation (TV) constraints are applied directly to the physical parameter maps.
- Self-Gating: Respiratory motion is handled via SSA-FARY, which extracts the breathing signal directly from the MRI data, allowing for motion-resolved reconstruction without external bellows.
The entire pipeline was implemented using the open-source Berkeley Advanced Reconstruction Toolbox (BART).
Validation and Datasets
The method was validated using a custom-built phantom and an in vivo cohort. The phantom consisted of eight tubes with varying concentrations of water, peanut oil, and iron nanoparticles to simulate a range of hepatic steatosis and iron overload states.
The in vivo study included 21 subjects (one healthy volunteer and 20 patients with obesity, diabetes, or suspected NAFLD). Results were compared against a reference breath-hold Cartesian multi-echo Dixon scan.
Results
In phantom experiments, the measured fat fraction showed excellent linear correlation ($y = 0.973x + 1.997$) with MR Spectroscopy. In the clinical cohort, the free-breathing radial maps demonstrated strong agreement with the breath-hold Cartesian references. Notably, in cases where patients failed to maintain a breath-hold during the Cartesian scan, the radial method produced superior, artifact-free $R_2^*$ maps. The technique also showed robustness to data undersampling; reducing the scan time by half maintained accurate mean parameter values, albeit with a slight increase in noise.
Discussion
A key advantage of this approach is the joint estimation of $B_0$ and coil maps, which accounts for field changes during respiration. The use of physics-based priors (e.g., spatial smoothness for $B_0$ maps) further improves reconstruction quality.
However, clinical implementation faces challenges. The reconstruction is computationally intensive, requiring approximately 4 hours on a high-end GPU for a full volume. Additionally, the automated frequency calibration failed in a subset of patients, requiring manual intervention. Future work will focus on multi-GPU parallelization and more robust automated initialization, as well as validation in patients with severe iron overload.
Conclusion
The proposed multi-echo radial technique provides a robust, free-breathing alternative for liver fat and iron quantification, potentially expanding the reach of quantitative MRI to patient populations unable to perform breath-holds.
Summary updated