A Physics-Aware Deep Neural Network for Real Time Volumetric MRI in Support of MR-Guided Radiotherapy
L Liu1*, L Shen1, A Johansson2, J Balter3, Y Cao3, L Xing1, (1) Stanford University, Palo Alto, CA, (2) Uppsala University, (3) University of Michigan, Ann Arbor, MI
Presentations
WE-C-TRACK 6-1 (Wednesday, 7/28/2021) 1:00 PM - 2:00 PM [Eastern Time (GMT-4)]
Purpose: To investigate the feasibility of real time volumetric MRI using a physics-aware deep neural network, in support of MRI-based 3D motion tracking for MR-guided radiotherapy.
Methods: A deep neural network was proposed to generate volumetric MRI from orthogonal Cine MRI, which can be acquired on MR-Linac systems real time. The network consisted of a 2D generation module that synthesized oblique slices from two orthogonal slices and a 3D refinement module to generate final high quality volumetric images. A physics module that utilized prior knowledge of imaging physics was used to bridge 2D and 3D domains. The proposed method was evaluated for 6 patients with intrahepatic carcinoma. Each patient received two golden-angle radial MRI examinations that were more than 1 month apart, where one examination was used for network training and another for evaluation. Time series of 100 volumetric images with a sampling rate of 340 ms were reconstructed via retrospective binning and served as the ground truth. Orthogonal Cine MRI were sampled from volumetric MRI and served as network inputs. A reference MRI with gross tumor volume (GTV) contours was deformed to match the time series. Deformed GTV centroid positions and contours were compared between ground truth and network prediction.
Results: Across the 6 patients evaluated, the averaged distance between predicted and ground truth GTV centroids was less than 1 mm in all 3 directions (anterior-posterior, inferior-superior and lateral). The averaged 95-percentile Hausdorff distances between predicted and ground truth GTV contours was 4.6 mm, which was similar to the cross-plane imaging resolution (4 mm).
Conclusion: Physics-aware deep neural network enabled real time volumetric MRI with sufficient accuracy for 3D motion tracking. The proposed technique is robust to longitudinal patient changes and has the potential of reducing treatment margins and improving treatment delivery precision on a MR-Linac system.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by NIH R01 EB016079.
Handouts
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email