Encrypted login | home

Program Information

Velocity-Based Reference Scan Selection for Motion-Model Error Reduction of 5DCT

L Yang

Liu (Lisa) Yang*, Dylan O'Connell , Tai Dou , David Thomas , James Lamb, Daniel A. Low, University of California, Los Angeles, Los Angeles, CA 90024


SU-E-J-156 (Sunday, July 12, 2015) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: A recently proposed 5DCT protocol uses fast helical CT scans to model lung tissue motion that requires one of the scans to be used as a reference for deformable image registration. We have found that, while the CT scans are acquired using fast helical acquisition, residual motion-artifacts remain for scans acquired during mid inspiration or exhalation. Selecting the scan with minimal motion artifacts as the reference scan will improve image registration and model accuracy.

Methods: A free-breathing patient was scanned 25 successive times in alternating directions with a Siemens 64-slice CT scanner using a low-mA fast-helical protocol. Each of the 25 CT scans was successively used as the reference scan with the other 24 images registered to it. Motion-model parameters were determined using the breathing-motion model. Voxel specific tissue velocity maps were generated for each of the 25 scan based on the breathing amplitude surrogate and its time derivative. The resulting 24 model error maps were registered to the first CT scan for comparison. The relationships between regional motion-model error, reference scan motion-artifacts and tissue-velocity were analyzed using linear-regression.

Results: Motion-model errors were linearly correlated with reference scan blurring artifacts and doubling artifacts. The effects of blurring-artifacts on motion model errors were stronger than doubling artifacts due to their impact on registration accuracy. Selecting the lowest tissue motion velocity scan as the reference scan decreased the regional motion model error by 33.67%±43.12%.

Conclusion: Motion-artifacts, especially blurring artifacts, that exist in the reference image induce motion model errors in 5DCT. A carefully selected reference scan based on the lowest tissue velocities can minimize the regional motion model error.

Contact Email: