抄録
Brain activation studies, pioneered by Sokoloff in 1961, are capable of providing "snap shots" of brain activities in terms of specific brain functionality and represent an ultimate tool for the localization based approach to understanding the human mind. Because of its completely non-invasive nature, fMRI, a technique based on T2* detectable MR signal alteration, reflecting a relative reduction of venous blood volume associated with increased regional blood flow in accordance with the Munro-Kellie doctrine, has quickly become the gold standard. A major draw back of fMRI, however, is its significantly high susceptibility to producing "fictitious activation." This is a common pitfall for investigators who lack proper understanding of MR specific signal properties. Use of linear statistical methods such as SPM, especially in highly demanding protocols, i.e. so-called event related fMRI, further accentuates the inherent weakness of fMRI and has resulted in the rather chaotic current situation in the world of brain science. The best remedy for this potential fatal problem is the application of statistical methods capable of automatically separating task-related activations from artificial components. ICS (see http://coe.bri.niigata-u.ac.jp) represents a well-validated, highly reliable technique in this category. The brain is a complex system and its function cannot be described as the liner sum of deterministic functional units. Therefore, absolute localizationism is fated for eventual failure. Nevertheless, because of its capability of performing single session multivariate analysis, ICS fMRI remains to be effective tool for non-linear dynamic analysis in brain science. [Jpn J Physiol 54 Suppl:S2 (2004)]
