Program Information
Identifying Abdominal Inter-Organ Motion Correlations Using 4D-MRI and 4D-Image Registration
Y Yue*, Z Fan , Z Deng , J Pang , J DeMarco , R Tuli , D Li , B Fraass , Cedars-Sinai Medical Center, Los Angeles, CA
Presentations
SU-F-303-9 (Sunday, July 12, 2015) 4:00 PM - 6:00 PM Room: 303
Purpose:
This study aims to identify motion correlations between multiple abdominal organs using high spatial resolution 4D-MRI images and 4D-image registration, and to model internal organ motion using known surface motions for individual patients.
Methods:
4D-MRI images were acquired at 3T using a spoiled gradient echo sequence with self-gated 3D projection reconstruction. Six healthy human subjects and a MRI motion phantom were scanned with the 4D-MRI imaging sequence. The images were reconstructed into 10 phases with isotropic spatial resolution [1.56 mm]^3. The 4D images were registered with an in-house 4D registration algorithm which uses the self-gating respiratory signals as the temporal regularization energy. A 4D deformation field was generated from the registration. In the human study, we examined motion of three regions of interest (ROIs), including abdominal skin, upper liver, and pancreas head. Each ROI contains 8,000-20,000 voxels. The primary motion trajectory was extracted by principle components analysis. The correlation of inter-organ motion was modeled by 3D orientation matching between motion trajectories. Liver and pancreas motions were projected onto the motion plane of abdominal surface resulting in rotation and scaling parameters. The correlation of inter-organ motions was evaluated using RMS errors of trajectory matching.
Results:
The accuracy of 4D-registration for the phantom data had RMS error 0.5±0.3mm. The human data registration accuracy, using manual tracking of landmarks, had RMS error 0.9±0.7mm. The motion ranges were 23±16mm (upper liver), 9.5±5mm (pancreas), and 10.2±6mm (surface). Compared to the surface motion, trajectories of liver and pancreas had 85±34 and 63±41 degree angle-of-motion and scale factors 2.4±1.3 and 1.5±1.1, respectively. The trajectory matching errors were 1.0±0.6mm (liver) and 0.7±0.5mm (pancreas).
Conclusion:
Abdominal inter-organ motion can be accurately characterized using 4D-MRI images and 4D-image registration techniques. The proposed method provides a feasible approach to more accurately use patient surface surrogates to estimate internal organ motions.
Contact Email: