Program Information
Novel Method for Geometrically Undistorted B0 Inhomogeneity Field Map Estimation and Image Correction
A Matakos*, J Balter , Y Cao , University of Michigan, Ann Arbor, MI
Presentations
WE-G-18C-1 Wednesday 4:30PM - 6:00PM Room: 18CPurpose: To develop and evaluate a method to estimate undistorted B0 inhomogeneity field maps of MRI and to correct susceptibility-induced geometrical and intensity distortions in clinical images in order to support precision RT planning.
Methods: The proposed method consisted of two major steps. First, undistorted B0 inhomogeneity field maps and images were estimated from k-space data of a 3D dual gradient-echo (GRE) acquisition using a novel joint reconstruction algorithm. After linearizing the cost function that includes two spatial regularizations, the algorithm iteratively estimated the undistorted field maps and images. Second, clinical images acquired with a low readout-bandwidth in the same imaging session, e.g., 180 Hz/pixel in clinical brain T1-weighted images, were corrected for geometrical and intensity distortions by using the undistorted field map. The method was evaluated through simulation, and studies of a geometric phantom with air-water interfaces and human brain on a 3T scanner.
Results: Simulation showed that our method reduced field map errors (RMS and maximum error) in the undistorted field maps ten-fold compared to the distorted ones, and geometric and intensity artifacts in the jointly reconstructed images. In the phantom study, the 2-mm distortion near the air-water interface was fully corrected using the undistorted field map, but only partially corrected using the distorted field map with approximately 1-mm residual distortion. In the patient study, there was >2 mm distortion near the sinus in the distorted field map. The geometrically accurate field map obtained by our method resulted in improved correction of geometry and intensity in both GRE and clinical T1-weighted images at the air/tissue interface.
Conclusion: Our method can correct geometrical distortion in clinical images, create a geometrical accurate image template, or provide a means for quality assurance of individual patient images, to enable MRI as a primary modality in precision RT planning.
Funding Support, Disclosures, and Conflict of Interest: Work supported by: NIH R01 EB016079
Contact Email: