The extreme variability in the structural conformation of the human brain p
oses significant challenges for the creation of population-based atlases. T
he ability to statistically and visually compare and contrast brain image d
ata from multiple individuals is essential to understanding normal variabil
ity within a particular population as well as differentiating normal from d
iseased populations. This paper introduces the application of probabilistic
atlases that describe specific subpopulations, measures their variability
and characterizes the structural differences between them. Utilizing data f
rom structural MRI, we have built atlases with defined coordinate systems c
reating a framework for mapping data from functional, histological and othe
r studies of the same population. This paper describes the basic approach a
nd a brief description of the underlying mathematical constructs that enabl
e the calculation of probabilistic atlases and examples of their results fr
om several different normal and diseased populations.