HyperSLICE: HyperBand optimised Spiral for Low-latency Interactive Cardiac Examination
share
BACKGROUND: Interactive cardiac magnetic resonance imaging is used for fast scan planning and MR guided interventions. However, the requirement for real-time acquisition and near real-time visualization constrains the achievable spatio-temporal resolution. PURPOSE: To improve interactive imaging resolution through optimization of undersampled spiral sampling and leveraging of deep learning for low-latency reconstruction (deep artifact suppression). POPULATION: Deep artefact suppression training data consisted of 692 breath-held CINEs. The developed interactive sequence was tested prospectively in 12 patients (10 for image evaluation, 2 during catheterization). ASSESSMENT: In simulated data, NRMSE, pSNR and SSIM of radial, uniform spiral and optimized spiral sampling were compared. In the prospective study, the optimized spiral interactive sequence was compared to conventional Cartesian real-time, and breath-hold cine imaging in terms quantitative and qualitative image metrics. RESULTS: The NRMSE, pSNR and SSIM were all statistically significantly higher in simulations of optimized spiral compared to radial and uniform spiral sampling, particularly after scan plan changes (SSIM: 0.71 vs 0.45 and 0.43). Prospectively, HyperSLICE proposed a higher spatial and temporal resolution than conventional Cartesian real-time imaging. The pipeline was demonstrated in patients during catheter pull back showing sufficiently fast reconstruction for interactive imaging. DATA CONCLUSION: HyperSLICE enables higher spatial and temporal interactive imaging. Optimizing the spiral sampling enabled better overall image quality and better handling of image transitions compared to radial and uniform spiral trajectories.
READ FULL TEXT