| Abstract: |
Objective. To develop motion-resolved volumetric MRI with 1.1 mm isotropic resolution and scan times <5 min using a combination of 3D radial kooshball acquisition and spatial-temporal deep learning 4D reconstruction for free-breathing high-definition (HD) lung MRI. Approach. Free-breathing lung MRI was conducted on eight healthy volunteers and ten patients with lung tumors on a 3 T MRI scanner using a 3D radial kooshball sequence with half-spoke (ultrashort echo time, UTE, TE = 0.12 ms) and full-spoke (T1-weighted, TE = 1.55 ms) acquisitions. Data were motion-sorted using amplitude-binning on a respiratory motion signal. Two high-definition Movienet (HD-Movienet) deep learning models were proposed to reconstruct 3D radial kooshball data: slice-by-slice reconstruction in the coronal orientation using 2D convolutional kernels (2D-based HD-Movienet) and reconstruction on blocks of eight coronal slices using 3D convolutional kernels (3D-based HD-Movienet). Two applications were considered: (a) anatomical imaging at expiration and inspiration with four motion states and a scan time of 2 min, and (b) dynamic motion imaging with 10 motion states and a scan time of 4 min. The training was performed using XD-GRASP 4D images reconstructed from 4.5 min and 6.5 min acquisitions as references. Main Results. 2D-based HD-Movienet achieved a reconstruction time of <6 s, significantly faster than the iterative XD-GRASP reconstruction (>10 min with GPU optimization) while maintaining comparable image quality to XD-GRASP with two extra minutes of scan time. The 3D-based HD-Movienet improved reconstruction quality at the expense of longer reconstruction times (<11 s). Significance. HD-Movienet demonstrates the feasibility of motion-resolved 4D MRI with isotropic 1.1 mm resolution and scan times of only 2 min for four motion states and 4 min for 10 motion states, marking a significant advancement in clinical free-breathing lung MRI. |
