A new look at finite mixture models in medical image analysis

Abstract

Presents a new look at finite mixture models in unsupervised medical image analysis. Both the conditional and the standard finite normal mixture models are discussed in detail in terms of physical and mathematical understanding. Based on statistics and information theory, their applications in model selection, parameter quantification and image segmentation are justified and supported by several new theorems and algorithms. Numerical examples with simulated data and real medical images are presented which have shown a great promise. >