Registration algorithms, applied in a probabilistic framework, offer a new method to examine abnormal brain structure. Probability maps can be combined with anatomically driven elastic transformations that associate homologous brain regions in an anatomic database. This provides the ability to perform morphometric comparisons and correlations in three dimensions between a given subject's MR scan and a population database, or between population subgroups stratified by clinical or demographic criteria.
Methods to compare probabilistic information on brain structure from different subpopulations are under rapid development. They include approaches based on random tensor fields [14,47,108,112,113], singular value decomposition and ManCova (multivariate analysis of covariance; ), shape-theoretic approaches , stochastic differential equations , and pattern theory . The resulting probabilistic systems show promise for encoding patterns of anatomic variation in large image databases, for pathology detection in individuals and groups, and for determining effects on brain structure of age, gender, handedness, and other demographic or genetic factors.
As well as disease-specific atlases reflecting brain structure in dementia and schizophrenia, research is underway to build dynamic brain atlases that retain probabilistic information on temporal rates of growth and regressive processes during brain development and degeneration. Refinement of these atlas systems to support dynamic and disease-specific data should generate an exciting framework to investigate variations in brain structure and function in large human populations.
Was this article helpful?