Fast bias field correction for 9.4 tesla magnetic resonance imaging

In this paper the problem of intensity inhomogeneity at high magnetic field on magnetic resonance images is addressed. Specifically, rat brain images at 9.4T acquired with a surface coil are bias corrected. We propose a low-pass frequency model that takes into account not only background-object contours but also other important contours inside the image. Two pre-processing filters are proposed: first, to create a volume of interest without contours, and second, to extrapolate the image values of such masked area to the whole image. Results are assessed quantitatively and visually in comparison to standard low pass filter approach, and they show as expected better accuracy in enhancing image intensity.