Nonparametric regression analysis of functional brain images

There has recently been remarkable progress in noninvasive measurement of the brain (such as f-MRI, MEG, and PET), providing detailed functional brain images. Using this technique, the correspondence between the areas and functions of the brain can be precisely analyzed. Researchers in this field are now examining correspondences between areas and functions of the brain on the basis of functional brain images. The authors have been attempting to apply nonparametric regression analysis to f-MRI images, in an effort to develop an algorithm that identifies the areas of the brain automatically and objectively. Since the brain has a three-dimensional structure, it is desirable to apply three-dimensional nonparametric regression analysis, but that requires a tremendous amount of computation. The idea is applied to two-dimensional images as a first step. Satisfactory results are obtained compared to the results obtained by z-score, a typical technique; the results are reported in this paper. © 2003 Wiley Periodicals, Inc. Syst Comp Jpn, 35(1): 67–78, 2004; Published online in Wiley InterScience (). DOI 10.1002sscj.10244