Encrypted login | home

Program Information

4D-MRI Reconstruction Using Group-Wise Registration

no image available
R Farah

R Farah1*, S Shea2 , E Tryggestad3 , R Teboh Forbang1 , J Wong1 , R Hales1 , J Lee1 , (1) Department of Radiation Oncology and Molecular Radiation Sciences, Johns Hopkins University School of Medicine (2) Department of Radiology, Loyola University Medical Center (3) Department of Radiation Oncology, Mayo Clinic

Presentations

SU-F-303-1 (Sunday, July 12, 2015) 4:00 PM - 6:00 PM Room: 303


Purpose:
The objective of this study is to develop an algorithm that accurately reconstructs a representative 4D-MRI from a multi-slice dynamic 2D-MRI scan.

Methods:
Multi-slice dynamic 2D-MR images at sagittal view are acquired using a cyclic and interleaved protocol to avoid crosstalk between adjacent slices and capture temporal anatomical variations. We chose sagittal view as it captures both superior-inferior and anterior-posterior directions where the most dominant respiration-induced anatomical motion happens. A respiratory signal is extracted directly from the sagittal MR images by computing the body area. The acquired images are first sorted by respiration phase-binning. To reconstruct each slice at each bin, we use group-wise-registration (GWR) on all the corresponding sorted images. GWR repeatedly registers the images to a template space. The template is initialized using a representative image that shows the best similarity to its adjacent slices among sorted images, and is updated at each iteration by averaging the resulting registered images. Finally, the reconstructed slices are stacked at each bin to form a 4D-MRI reconstruction.

Results:
We tested our method on five lung cancer patient data sets scanned on a Siemens 1.5T MAGNETOM Espree scanner using a balanced steady-state free precession sequence (TrueFISP, TR=3 ms, TE=1.22 ms, flip angle=77-79°). We compared our method with two conventional slice average-based methods. Our method has significantly enhanced the reconstruction quality as well as the edges, and showed improved edge sharpness by an average of 20% (in terms of the frame average strength response to the Deriche edge operator) compared to conventional methods.

Conclusion:
We proposed a new 4D-MRI reconstruction algorithm that enhances the 4D-MRI reconstruction quality over the conventional methods. We initially tested this method on lung cancer cases, but it can be extended to any sites where respiration-induced motion is critical.

Funding Support, Disclosures, and Conflict of Interest: Funding From an NIH grant NIH/NCI 5R21CA178455


Contact Email: