Encrypted login | home

Program Information

A Preliminary Evaluation of Respiratory Motion Modeling at 0.35T for MRI-Guided Radiotherapy

J Ginn

J Ginn1*, D O'Connell1 , D Thomas2 , N Agazaryan1 , M Cao1 , Y Yang1 , D Low1 , J Lamb1 , (1) University of California, Los Angeles, Los Angeles, CA, (2) University of Colorado, Denver, CO


TU-FG-FS2-2 (Tuesday, August 1, 2017) 1:45 PM - 3:45 PM Room: Four Seasons 2

Purpose: To build a respiratory motion model based on 0.35T MRI imaging. The model estimates tissue motion using a respiratory surrogate, allowing a high signal to noise ratio 3D image to be reconstructed at any breathing phase. The model may be used in the future to estimate accumulated dose during a radiotherapy fraction subject to respiratory motion, and/or extract local measures of tissue properties.

Methods: A balanced steady state free precession sequence was used to acquire volumetric sets of interleaved sagittal 2D images at 3 frames/sec during free breathing, as well as a 3D breath hold reference volume. An external respiratory surrogate was recorded during imaging using a respiratory bellows. Tissue motion vectors derived from 2D deformable registration of the 2D images to the reference volume were fit to a linear model relating the amplitude and derivative of the surrogate to tissue motion. To evaluate model stability, the 2D images were partitioned into two sets, two models were extracted and the estimated tissue positions were compared. Model error was estimated using a cross-validation approach by omitting a single frame from model fitting, and comparing the model-estimated to the actual tissue location.

Results: The median and standard deviation tissue location difference predicted by the two models extracted from the two data partitions were 1.05 mm and 1.93 mm respectively. The differences in estimated tissue positions from these two models were less than 1.0 mm 2.0 mm and 3.0 mm for 47.8%, 76.8% and 89.2% of the voxels. The median and standard deviation cross-validation estimated errors were found to be 2.76 mm and 1.86 mm respectively.

Conclusion: A proof of concept 0.35T MRI-based linear respiratory motion model was demonstrated. This technique requires further refinement and must be extended to include 3D deformation before it can be used for dose calculations during radiotherapy.

Funding Support, Disclosures, and Conflict of Interest: Some authors have received consulting fees and speaking fees from ViewRay.

Contact Email: