Deep learning within a priori temporal feature spaces for large-scale dynamic MR image reconstruction: Application to 5-D cardiac MR Multitasking
This work addresses the problem of slow reconstruction times for high-resolution dynamic MRI in clinical settings, making it more practical for routine use.
The authors tackled the challenge of large-scale dynamic MRI reconstruction by applying deep neural networks within a priori temporal feature spaces, enabling global temporal modeling for sequences with over 40,000 frames and achieving a 3000x speedup in feature space calculation, from 20 minutes to 0.39 seconds.
High spatiotemporal resolution dynamic magnetic resonance imaging (MRI) is a powerful clinical tool for imaging moving structures as well as to reveal and quantify other physical and physiological dynamics. The low speed of MRI necessitates acceleration methods such as deep learning reconstruction from under-sampled data. However, the massive size of many dynamic MRI problems prevents deep learning networks from directly exploiting global temporal relationships. In this work, we show that by applying deep neural networks inside a priori calculated temporal feature spaces, we enable deep learning reconstruction with global temporal modeling even for image sequences with >40,000 frames. One proposed variation of our approach using dilated multi-level Densely Connected Network (mDCN) speeds up feature space coordinate calculation by 3000x compared to conventional iterative methods, from 20 minutes to 0.39 seconds. Thus, the combination of low-rank tensor and deep learning models not only makes large-scale dynamic MRI feasible but also practical for routine clinical application.