Cine-DL: Deep learning framework enables motion-robust free-breathing cardiac MRI

Conventional cardiac cine MRI relies on breath-hold Cartesian acquisitions, which are vulnerable to motion artifacts and can be uncomfortable or infeasible for pediatric and noncompliant patients. Free-breathing radial acquisitions alleviate these limitations but suffer from prominent streak artifacts at high acceleration. To address this, Mahmut Yurt et al. propose Cine-DL, a clinically oriented deep reconstruction framework. The pipeline first performs retrospective cardiac binning and respiratory gating on raw free-breathing radial data to resolve cardiac phases and discard motion-corrupted spokes. Then, Streak Optimized Coil Compression (SOC) explicitly preserves cardiac signals while suppressing peripheral interference that drives streak artifacts. The resulting 2D+t cine series is reconstructed using an unrolled network that alternates a ResNet proximal operator with physics-based data consistency updates solved via conjugate gradient. A memory-efficient training strategy reduces peak memory usage.

Cine-DL was evaluated on free-breathing volunteer data against established baselines (k-t SENSE and iGRASP) and demonstrated clinical translation on newly acquired patient data. The framework consistently improved quantitative metrics and visual fidelity, producing high-quality cardiac cine images without requiring breath-holding. The authors note that this work supports a practical route toward routine, time-sensitive clinical adoption of free-breathing cine MRI. By combining targeted preprocessing with model-based deep learning, Cine-DL offers a robust solution for motion-robust cardiac imaging, particularly benefiting patients who cannot reliably hold their breath.