The novel intracranial boundary detection method developed in Chapter 7 of this thesis produces apparently accurate intracranial contours for the 5 data sets presented in this chapter. Except for the entry of MRI-specific parameters, such as size and resolution, the method required no user interaction. Expert evaluation of the method remains as future work. However, the brain masks produced by the method have been used successfully by the Correct Intensity process (also developed in Chapter 7).
Results show that the intensity variation due to RF inhomogeneity is reduced in the 3 data sets corrected by the Correct Intensity process. The impact of this reduction, measured by the success of Johnston's MS lesion segmentation algorithm, is currently under investigation [24].