Capturing Breathing Variability Using Surface Tracking-assisted Time-resolved Multi-cycle 4D Lung MRI.

Abstract

Objective: To develop time-resolved, multi-cycle, MRI (TRMC-MRI) for 4D lung imaging that can capture respiration-induced cycle-to-cycle variations in the internal anatomy. Approach: Golden-angle 3D stack-of-stars gradient echo data were continuously acquired during free breathing for 2 minutes, or 4000 radial views. Thoracoabdominal surface motion was concurrently tracked by four MR-compatible electromagnetic motion tracking sensors at a frequency of one tracking event for every two radial views. A radial view was a stack of k-space radial spokes acquired with the same radial angle for all the partition phase encoding steps. A continuous breathing state was defined for each tracking event, and the two radial views associated with each tracking event, by the principal component scores based on the sensor positions. To reconstruct a dynamic volume for a tracking event, radial views with similar breathing states to the tracking event in question were collected from the entire acquisition to fill the k-space. Sensitivity maps were estimated from all the acquired radial views. Reconstruction of dynamic volumes was performed with parallel imaging with total variation regularization. The proposed method was performed on four healthy volunteers (Male/Female: 3/1, Age: 30±2.3 years) in the right lung. Main results: Thoracoabdominal surface tracking showed cycle-to-cycle breathing variability (coefficients of variation for period: 8-23%, for amplitude: 7-36%) despite instruction of breathing regularly. Dynamic lung volumes covering 320x320x24mm3 were generated at every 60.6ms for the entire 2-minute acquisition consisting of on average 23.2 (range: 18-34) breathing cycles. Considerable breathing variations were captured in time-resolved multi-cycle breathing motion. Significance: The surface tracking-assisted TRMC-MRI framework can provide critical breathing variations information for MR-guided lung radiotherapy, including treatment planning, motion modeling and prediction, and training for real-time MR in the treatment room.