A review of the behavior of different diffusion anisotropy indices (DAIs) to changes in diffusion anisotropy was undertaken and the limitations of several previously proposed DAIs are discussed. A mathematical representation of anisotropy sensitivity was used to enable image contrast and sensitivity to noise to be predicted. This technique was utilized to devise a new DAI, the gamma-variate anisotropy, which has a low sensitivity to noise and an improved contrast range for human brain imaging. This index can be easily modified for other applications, in that the range of anisotropy values to which it is highly sensitive can be altered. (C) 2000 Wiley-Liss, Inc.