Premium
Poster — Wed Eve—50: Correcting for Fat‐Shift Artifacts in Magnetic Resonance Images
Author(s) -
Baldwin LN,
Wachowicz K,
Gino Fallone BG
Publication year - 2009
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 - 10.1118/1.3244154
Subject(s) - distortion (music) , residual , signal (programming language) , transformation (genetics) , artificial intelligence , artifact (error) , computer vision , image processing , computer science , image (mathematics) , algorithm , mathematics , chemistry , telecommunications , amplifier , biochemistry , bandwidth (computing) , gene , programming language
The inclusion of magnetic resonance (MR) images in radiation therapy treatment planning has been hampered by inherent image distortions. As such, distortion correction algorithms have been under investigation for many years. Although distortion maps can be derived through a number of methods, a common problem arises when the resultant distortion transformation is not unique. This situation is present at the interfaces between tissues experiencing different chemical shift environments (i.e. at fat / water boundaries), and leads to hypo‐ and hyper‐intense artifacts in the distorted image. Correcting image distortion by interpolating between true and distorted coordinates cannot fully correct for the erroneous high intensity region because the distortion transformation at such locations is not unique. Thus, residual distortion artifacts remain and may hinder the accurate delineation of areas associated with fat signal, i.e. external body contours. Here we propose a series of image processing steps which are carried out prior to the standard distortion correction procedure. The method involves establishing regions of signal overlap in the distorted image (i.e. where the distortion transformation is not unique) and separating the resultant high signal into the water‐based signal component and the shifted fat‐based signal component. The fat signal is unshifted prior to the standard distortion correction technique. When the described methods are applied, the residual high‐intensity regions associated with the chemical shift artifact can be eliminated.