JACK KROHMER EARLY-CAREER INVESTIGATOR COMPETITION WINNER: A Time-Resolved Slice-to-Volume Reconstruction Framework with Dynamic Reconstruction and Motion Estimation (S2V-DREME)
Abstract
Purpose
Existing volumetric MRI techniques are constrained by the trade-off between acquisition time and image quality, limiting accuracy in motion-impacted sites such as the liver. To enable fast and high-quality volumetric imaging with sufficient spatiotemporal resolution, we developed a time-resolved volumetric MRI technique that recovers 3D volumes from two orthogonally acquired MR slices for real-time 3D tracking.
Methods
Two orthogonal 2D slices, dynamically acquired in time and space, were mapped to time-resolved 3D MRIs using a one-shot slice-to-volume framework, S2V-DREME. The model jointly estimates a reference 3D MRI and time-resolved deformation-vector-fields (DVFs) that warp the reference volume into dynamic 3D MRIs. The reference volume is represented by a spatial implicit-neural-representation (INR), while the DVFs are derived via low-rank motion modeling. Motion basis components (MBCs) are generated by a spline-enhanced INR (SINR)-based motion generator, and corresponding coefficients are inferred via a feature-wise linear modulation (FiLM)-based orthogonal-slice motion encoder. A progressive optimization strategy is employed, in which the spatial INR and MBCs are initialized sequentially before joint optimization. The overall loss function combines 2D-slice data fidelity, total variation (TV) regularization, and an MBC normalization. S2V-DREME was evaluated on digital phantom (XCAT) simulations, physical phantom measurements, and human studies.
Results
S2V-DREME generates high-resolution volumetric MRIs by fusing the high in-plane resolution from two complementary orthogonal views. For XCAT, it captures regular/irregular motion accurately for both dynamic reconstruction (training phase, mean DSC/COME: 0.9251±0.0308 / 0.98±0.43mm) and real-time motion estimation (testing, mean DSC/COME: 0.9125±0.0164 / 0.99±0.73mm). Physical phantom experiments achieved a mean COME of 1.14±0.05mm, and human studies confirmed high-resolution 3D reconstruction from each orthogonal slice pair.
Conclusion
By combining a novel step-and-shoot acquisition protocol with motion-compensated one-shot learning, S2V-DREME enables accurate time-resolved volumetric MRI reconstruction and motion tracking from cine orthogonal-view slices, with strong potential for rapid volumetric imaging and real-time MR-guided adaptive radiotherapy.