Fifty-seventh annual meeting of the American association of physicists in medicine
TU-CD-BRA-06: 3D Reconstruction From 2D CineMRI Orthogonal Slices: A Feasibility Study
The ideal image guidance strategy to acquire complete spatio-temporal information for the patients during therapy. MRI-Linacs come close to this ideal, but there are spatial and temporal trade-offs. Several authors have proposed simultaneous orthogonal images as a guidance strategy for MRI-Linac. In order to estimate the missing 3D information from the 2D orthogonal images we present a feasibility study on the reconstruction of a 3D volume directly from 2D orthogonal CineMRI slices for MRI-guided treatment.
A 4DCT XCAT lung tumor phantom was generated in two respiratory phases (i.e. exhale and inhale) with 3cm diaphragm motion. We simulated the cine acquisition by extracting sagittal and coronal orthogonal slices from the inhale XCAT phantom. Then, we performed DIR between the selected slices at the inhale phase with the corresponding ones at exhale. An interpolation of the derived motion field was performed in the RL, AP and SI directions, thus obtaining a 3D motion field that was used to warp the exhale phase and obtain a 3D estimation of the inhale phase (i.e. estimated_inhale volume). As validation we compared the estimated_inhale volume with the inhale phase (i.e. ground truth). We also applied the method on one CineMRI lung patient image set, where inter-leaved sagittal/coronal cineMRI images were available.
The estimated_inhale volume resulted comparable with the ground truth. The distance between landmarks identified in the two respiratory phases was quantified as 24.8±7.2mm for exhale/estimated _inhale, vs.21.9±9.8mm for exhale/inhale. Tumor motion between exhale and estimated_inhale was 28.3mm, consistent with the measured diaphragm motion (i.e. 30mm). The patient results confirmed that the estimated_inhale volume compensated for organ motion (i.e. diaphragm distance between inhale and estimated_inhale was within 4.7±2.4mm).
The method provided a framework for the volumetric reconstruction from 2D orthogonal slices. The application to a patient case showed the capability to account for organ motion.