Maria Murgasova1, Joseph V. Hajnal2, Serena J. Counsell2, A David Edwards2, Daniel Rueckert1
1Department of Computing, Imperial College London, London, UK; 2Robert Steiner Magnetic Resonance Unit, Imaging Sciences Department, MRC Clinical Services Centre, Imperial College London, London, UK
We propose a novel template-based bias correction method highly effective for removal of stronger bias fields usually present in MR images acquired by modern high magnetic field MR scanners. The bias field is estimated by weighted least square B-spline fitting to a difference of log-transformed image and an aligned and intensity-adjusted template image. The method is applied 35 pediatric brain MR images acquired using a MP RAGE sequence on a 3T MR scanner. The bias correction with B-spline control point spacing between 50 and 75 mm substantially improves EM-based classification of brain tissues in these images.