A gaussian mixture model approach for estimating and comparing the shapes of distributions of neuroimaging data: diffusion-measured aging effects in brain white matter.

Neuroimaging signal intensity measures underlying physiology at each voxel unit. The brain-wide distribution of signal intensities may be used to assess gross brain abnormality. To compare distributions of brain image data between groups, t-tests are widely applied. This approach, however, only compares group means and fails to consider the shapes of the distributions. We propose a simple approach for estimating both subject- and group-level density functions based on the framework of Gaussian mixture modeling, with mixture probabilities that are testable between groups. We demonstrate this approach by application to the analysis of fractional anisotropy image data for assessment of aging effects in white matter.