Title: Four-Dimensional Diffusion-Weighted Magnetic Resonance Imaging (4D-DWI): Development and Evaluation
Abstract: Diffusion-Weighted MR Imaging (DWI) provides superior tumor-to-normal-tissue contrast compared with techniques currently used for tumor motion tracking, including non-contrast CT and T2-weighed MRI. This study aims to develop a novel 4D-DWI technique using a hybrid sorting algorithm for imaging respiratory motion for radiation therapy. Image acquisition of 4D-DWI was achieved by repeatedly imaging a volume of interest using sequential-mode 2D DWI sequence in axial planes. To determine the number of repetitions (NR) needed to assure complete data completion (ie, that images of all respiratory phases are acquired for all slices), we performed computer simulations using Real-time Position Management (RPM) respiratory signals of 29 cancer patients to derive the relationship between NR and influencing factors, including number of slices scanned (NS), number of respiratory phases of the 4D-DWI (NP), and starting phase at image acquisition (P0). A frame rate of ∼2 frames/s was assumed for DWI image acquisition in the simulation. Retrospective sorting and reconstruction of 4D-DWI were achieved using a novel, hybrid (phase and amplitude) sorting algorithm utilizing redundantly acquired images. The techniques for 4D-DWI acquisition and reconstruction were evaluated using both a 4D digital human phantom (XCAT) via computer simulation and healthy volunteer cases under an IRB protocol. Percentage of complete acquisition of all respiratory phases (Cp) increased as NR increased in an inverse-exponential fashion (Cp=100*[1-exp (-0.28*NR)], when NS=30, NP=6). The NR needed for 4D-DWI (defined as achieving 95% completion, Cp=95%) showed a linear relationship with NP (NR∼1.8xNP, r=1.0), but was independent of NS and P0. Simulated 4D-DWI on the XCAT phantom showed clear patterns of respiratory motion, consistent with the input signal: the cross-correlation coefficient (CC) and the mean relative difference in motion amplitude (D) were 0.97 and 1.5%, respectively, in the superior-inferior direction, and 0.99 and 0.04% in the anterior-posterior direction. The 4D-DWI of healthy volunteers also revealed respiratory motion of internal organs. The mean relative difference in motion trajectory of a selected region of interest (left kidney) was 2.8% between 4D-DWI and single slice sagittal cine MRI, which was used as reference. A novel 4D-DWI technique has been developed and validated. This has significant potential to improve the visualization and delineation of mobile cancers for radiation therapy.