Anatomical atlases, traditionally in the form of books with images of 2-D sections and 3-D drawings, play a crucial role in the use of a standardized nomenclature to understand and describe anatomy and function of the human body. Early atlases mostly used dissections and labeling of single individuals, thus displaying the anatomy of one specific subject. With availability of digital 3-D imaging, computational anatomy tools and progress from statistical shape analysis, the scientific community got access to new methodologies to build digital atlases that represent the average anatomy of a population but also incorporate information about variability, often represented via deformation maps. Such digital atlas image data, together with expert labeling of anatomical and functional regions, are commonly used as spatial priors for Bayesian-type segmentation or to automatically label image data of new subjects via co-registration.
With age-related changes as in infant growth and degeneration due to aging, a single normative atlas will not properly represent expected anatomy and appearance of image data. Moreover, in the presence of pathology, the use of statistical atlases may fail since co-registration between healthy and abnormal may be ill-defined since pathology signatures and their localities are not coded in the atlases.
This talk will discuss the development of longitudinal statistical atlases that capture expected geometry and its variability due to age changes. We will demonstrate that statistical concepts of longitudinal data analysis such as linear and nonlinear mixed-effect modeling, commonly applied to univariate or low-dimensional data, can be extended to structures and shapes modeled from longitudinal image data. With pathology defined as differences from the expected norm, including geometric deformations and appearance changes, the use of normative atlases will be extended for automatic quantification of structures not described by the norm. This is achieved by a joint registration-segmentation method for classification of normal appearing and injured brain tissue.
The talk will discuss results from ongoing clinical studies such as analysis of early brain growth in subjects at risk for mental illness, early prediagnostic brain changes in autism, analysis of neurodegeneration in normal aging and Huntington's disease, and quantitative assessment of recovery in severe TBI. However, presented methodologies are generic in regard to uses in other application domains.
