Qualitatively, the new intracranial boundary detection technique proved accurate in tests on 5 MRI data sets from 2 different MRI scanners. This new technique exceeds the capabilities of previously suggested techniques in that it finds the intracranial boundary in all slices of the MRI scans without user interaction and in the presence of intensity variation due to RF inhomogeneity. Although quantitative analysis remains incomplete because of a lack of expert data from a radiologist, the automatically detected intracranial boundaries have been successfully used to isolate the brain for RF correction.
The proposed RF correction scheme quantitatively reduced unwanted low frequency intensity variation in the 3 MRI scans on which it was tested. Visual inspection of the corrected scans revealed that image detail was not lost as a result. Such correction has been shown to improve the results of automated tissue segmentation techniques [14][26].