Automatic intracranial boundary detection is important for brain tissue segmentation, for multimodal image registration, and for teleradiology and disk storage where lossy image compression is used. RF correction is vital to the success of computer-automated tissue segmentation techniques.
A novel method for fully automatic intracranial boundary detection and RF correction in MRI scans of the head was developed in this thesis. The new method consists of four steps:
Using histogram analysis, the Segment Head step distinguishes the head from background noise. The Generate Initial Brain Mask step isolates the brain in the segmented head via nonlinear anisotropic diffusion and automatic thresholding. With the initial brain mask as a seed, the Generate Final Brain Mask step uses active contour models to detect the intracranial boundary. In the final step, Correct Intensity, RF correction is performed on the masked brain with a homomorphic filter.