Editorial Multivariate brain mapping in clinical neuroscience research Over the past decade a number of innovative develop- ments in imaging technology have expanded the frontiers of clinical investigation in the neurosciences. Radiotracer imaging with positron emission tomography (PET) and sin- gle photon emission computerized tomography (SPECT) have become increasingly available for patient-oriented re- search. Moreover, highly innovative magnetic resonance imaging (MRI) techniques have been developed to provide critical information regarding structure/function relation- ships in the brain in both health and disease. In this regard, minimally invasive imaging techniques to assess neural func- tion have recently found application in natural history stud- ies and in clinical trials of novel therapies for brain disease. Although these uses of functional imaging have focused mainly on the mapping of hemodynamic responses during cognitive or pharmacological activation, resting state mea- surements of regional cerebral blood volume and flow, and of regional metabolism may provide a simple means of assessing disease processes in large study populations. The practical utility of these functional imaging ap- proaches in the clinical neurosciences has been enhanced by parallel developments in the field of image analysis. Computational advances have allowed for the implementa- tion of fast, unbiased software routines performed using either predefined volumes of interest from probabilistic at- lases, or voxel-by-voxel searches of the entire brain trans- formed into a standard anatomical space. Univariate methods as those employed in statistical parametric map- ping (SPM), can detect localized changes in discrete ana- tomical areas. These methods generally do not address interactions among brain regions. By contrast, multivariate approaches are designed to identify functional covariance patterns that correspond to spatially distributed neural sys- tems. We have found multivariate approaches to be partic- ularly useful in the study of neurodegenerative diseases in which dysfunction is evident not only at the site of pathol- ogy, but also in spatially remote regions that are elements of associated neural systems [1]. An important advantage of this approach has been the ability to quantify the expres- sion of an established disease-related pattern in individuals scanned prospectively in new populations. In effect, this attribute reduces the complex multi-dimensional image data from a subject into a specific score that can be used to predict the functional state of that individual under var- ious experimental conditions. Principal component analysis (PCA) belongs to a general class of multivariate statistical approaches designed to ana- lyze functional connectivity in large imaging datasets [2–4]. PCA applications have typically focused on the covariance mapping of regional brain function in the rest state, as well as that of activation responses mapped with PET or fMRI during task performance. Using classical cross-sectional ap- proaches, investigators have identified specific spatial covariance patterns associated with healthy aging [5,6], as well as with neuropsychiatric conditions such as Parkinson’s disease (PD) [7–9], Huntington’s disease (HD) [10,11], tor- sion dystonia [12], Alzheimer’s disease (AD) [13], vascular dementia [14], and mild cognitive impairment (MCI) [15]. Advances in PCA algorithms for within-subject designs, as used in activation studies, have led to the delineation of net- works linked to task performance in healthy volunteers and in patients with neurodegenerative disease [16,17]. Impor- tantly, such ‘‘signature patterns’’ have been found to have excellent test-retest reproducibility [9,18]. Moreover, these abnormal metabolic networks have been demonstrated to progress with time [15,19] and their expression in patients can be improved by treatment [20]. The recent demonstra- tion that PCA strategies are applicable to anatomical MRI data [21,22] represents a major advance in the clinical inves- tigation of brain disease with network biomarkers. The current issue of Clinical Neuroscience Research reviews the latest applications of network analysis in the study of neuropsychiatric diseases with functional brain imaging. In the first paper, Eckert and Edwards showcase a series of PCA-derived metabolic brain networks related to the parkinsonian syndromes and summarize the clinical use of network quantification in improving the accuracy of differential diagnosis in these disorders. The second paper by Huang, Mattis, and Julin describes specific functional brain networks associated with MCI syndromes in AD and PD. The authors show how network activity combined with psychometric data can increase the specificity and sen- sitivity of AD diagnosis in the prodromal MCI period. The paper by Carbon, Feigin, and Eidelberg focuses on the identification of genotype-related functional networks in clinically healthy carriers of mutations for autosomal 1566-2772/$ - see front matter Ó 2007 Association for research in Nervous and Mental Disease. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.cnr.2007.08.001 www.elsevier.com/locate/clires Clinical Neuroscience Research 6 (2007) 357–358