Premium
SU‐E‐J‐60: In Vivo Management of Suspectibility‐Induced Distortions for Intracranial MRI Guided Therapies at 3 Tesla
Author(s) -
Hwang K,
Stefan W,
Stafford R
Publication year - 2013
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4814272
Subject(s) - distortion (music) , image resolution , scanner , artificial intelligence , computer science , bandwidth (computing) , pixel , voxel , computer vision , physics , optics , mathematics , telecommunications , amplifier
Purpose: To assess, in vivo, subject‐dependent susceptibility‐induced shifts along the frequency encode direction, and develop a strategy for correcting such distortions for applications in need of high spatial accuracy. Methods: Distortion maps of the brain were created for 5 volunteers using data from a 16‐echo multiple gradient echo (MFGRE) sequence, acquired on a 3 Tesla scanner for R2* mapping applications. Both an R2* map and a proton resonance frequency (PRF) map was calculated using a single peak autoregressive moving average (ARMA) model, and aliased frequencies were unwrapped with a compressed sensing type non‐convex optimization method, using the assumption that sharp changes in the field are sparse. A map of susceptibility induced shifts is then calculated with knowledge of the frequency gradient, which scales with field of view and receiver bandwidth. The need for distortion correction is then determined either visually or automatically based on a threshold. If distortion correction is necessary, a smoothed version of the distortion map can be corrected to subvoxel resolution using conventional image deformation techniques. Results: Due to the high frequency encoding gradient of MFGRE sequences (±125 kHz over 25.6 cm FOV in our protocol), susceptibility‐induced distortions within the brain were less than 0.7 mm. However, the distortion map and correction method was applied to sequences with lower bandwidth in the same slice locations. Most distortions occurred in high susceptibility areas near the sinuses, with the optic chiasm shifting by approximately one pixel on low bandwidth, high resolution images. Conclusion: The creation of field and distortion maps for assessment and correction of distortion was performed using data from an MFGRE sequence. The additional R2* information from the same sequence provides an added benefit over dedicated field mapping sequences. With proper design of imaging protocols, distortions in the brain can be adequately contained for therapy planning purposes. Ken‐Pin Hwang is an employee of General Electric Healthcare.