The widespread use of neuroimaging methods in a variety of clinical and basic science fields has created the need for systematic and highly automated image analysis methodologies that extract pertinent information from images in a way that enables comparisons across different studies, laboratories, and image databases. Quantifying morphological characteristics of the brain from tomographic images, most often magnetic resonance images (MRIs), is an invaluable way for researchers to understand the way in which a disease can affect brain anatomy, to construct new diagnostic methods utilizing image information, and to create longitudinal follow-up studies evaluating potential drugs. The conventional type of morphological analysis of brain images has relied on manual tracings of regions of interest (ROI) (Bobinski et al., 1999; Convit et al., 1997; Cuenod et al., 1993; De Santi et al., 2001; deToledo-Morrell et al., 1997; Dickerson et al., 2001; Du et al., 2001; Frisoni et al., 1996; Jack et al., 1997, 1999; Killiany et al., 1993, 2000; Krasuski et al., 1998; Laakso et al., 1995, 2000; Lehericy et al., 1994; Rosen et al., 2003; Xu et al., 2000). These methods typically require that the reliability and repeatability of manual tracings be established across different raters, but also within the same rater at different times, as a first step. However, methods based on manually defined ROIs are limited in many ways. First, they rely on the need for a priori knowledge of the regions that are affected by a disease, so that respective ROIs can be defined, and therefore they may fail to discover new findings. Although a good hypothesis might be available at the beginning of a morphometric study, one would typically want to discover new knowledge that cannot, by definition, be part of the hypothesis. Take one example selected from the literature on the neuroimaging of dementia: although the role of hippocampal and entorhinal cortical atrophy in early prediction of Alzheimer's disease (AD) is widely accepted, relatively little is known about the potential involvement of other brain regions. Such knowledge could help in the construction of more sensitive methods for detection of and differentiation among different dementias. The complete investigation of the role of all brain structures in a disease, as well as the diagnosis of that disease, would be prohibitively labor intensive for an adequately large sample size if manual methods were employed. Moreover, inter- and intra-rater reliability issues would become crucial limiting factors, particularly in longitudinal studies, in which it is extremely difficult to maintain intra- and inter-rater reliability over time. Second, the spatial specificity of ROI-based methods is limited by the sizes of the ROIs being measured, which are typically rather coarse. A region that might be affected by disease may be only part of one predefined ROI, or it might span two or more ROIs, which inevitably washes out the results and reduces the statistical power of the measurement method. Alternative methods, such as stereology, are limited in a similar way. Although, in principle, one could delimit the ROIs measured to be as small as desired in order to increase spatial specificity, this would decrease rater reliability for measurement methods that are based on human raters. Finally, manual ROI tracing is severely limited in many modern studies, since it is no longer unusual to include over a thousand scans per study.