Inline SIMBA reconstruction for FISS whole-heart imaging

Background: Free-running whole-heart MRI provides free-breathing motion-resolved 5D cardiac imaging [1] without ECG triggering or respiratory gating, offering an operator-friendly approach to cardiovascular MRI (CMR). Fast interrupted steady-state (FISS) imaging [2], improves fat suppression and robustness to off-resonance artefacts compared to native bSSFP providing a useful platform for uninterrupted free-running whole-heart imaging. However, the computationally demanding 5D image reconstruction poses a barrier for inline application. The goal of this study is to evaluate the feasibility of the similarity-driven multidimensional binning algorithm (SIMBA [3,4]) for a fast static 3D image reconstruction directly on the MRI scanner for a free-running FISS acquisition.

Methods: SIMBA compensates motion by clustering periodically acquired superior-inferior projections based on similarity and forwarding only the 3D radial shots associated with the largest cluster to a 3D gridded reconstruction. Two subjects were scanned, one with and one without gadolinium using a prototype FISS sequence [2] (res:1.3mm³, TA:6:18, TR/TE:3.0ms/1.5ms, 3D phyllotaxis trajectory [5]). SIMBA (I_SIMBA) reconstructions were performed inline on a 1.5T clinical MRI system (MAGNETOM Sola, Siemens Healthineers, Forchheim, Germany) and the reconstruction time was measured. As references, data were also reconstructed with non-motion-suppressed regridding (I_AllData) and with the 5D free-running framework [1,6,7]. As a measure for the motion suppression, I_SIMBA and I_AllData were compared to the end-expiration mid-diastolic 5D image (I_5D) using the structural similarity (SSIM) index, assuming a higher SSIM correlates with improved motion suppression.

Results: SIMBA reconstructions of FISS datasets were successfully generated inline on the MRI system within 3min. Compared to I_AllData, SIMBA reduced motion blurring for FISS acquisitions with (Figure 1) and without (Figure 2) gadolinium. I_SIMBA provides sufficient image quality to visualize structures that would otherwise be affected by motion blurring (e.g., the right ventricle as shown in Figure 1). No major residual artefacts were observed, although mild blurring persisted in a subset of regions, like parts of the posterior myocardial wall. Quantitative analysis demonstrated that I_SIMBA show a higher motion suppression compared to I_AllData, shown by the SSIM_5D-SIMBA of 0.63±0.06 compared to SSIM_{5D-All_data} of 0.59±0.03.

Conclusion: This study demonstrates the feasibility of the first inline implementation of the SIMBA reconstruction for a FISS whole-heart acquisition. By enabling 3D motion-compensated reconstruction directly on the MRI system within clinically feasible times, SIMBA paves the way for clinical translation of free-running 5D MRI, supporting robust CMR without ECG or respiratory gating. Future work will focus on larger patient cohort studies, and integration into clinical workflows to further enhance accessibility of advanced cardiac MRI.

Qualitative comparison of the end-expiratory mid-diastolic phase of the 5D reconstruction (left), SIMBA reconstruction (middle) and non-motion-compensated “All data” image (right) for a FISS acquisition with gadolinium contrast enhancement. The yellow arrows highlight the areas where the SIMBA results in stronger motion suppression compared to the “All data” image.
Qualitative comparison of the end-expiratory mid-diastolic phase of the 5D reconstruction (left), SIMBA reconstruction (middle) and non-motion-compensated “All data” image (right) for a FISS acquisition without contrast enhancement. The yellow arrows highlight the areas where the SIMBA results in stronger motion suppression compared to the “All data” image.