Factor Analysis of Medical Image Sequences (FAMIS): Fundamental principles and applications

Abstract

Factor Analysis of Medical Image Sequences aims at resolving a medical image sequences into its underlying fundamental functions and their associated fundamental spatial distributions, yielding a description of the underlying physiological processes. We describe new developments regarding three stages of FAMIS. The clustering step is solved using an original method combining the criteria of spatial contiguity, signal evolution similarity, and the rule of the mutual nearest neighbours. A statistical model for medical image sequences, the fixed effect model, applied to medical image sequences, gives a theoretical basis to the choice of the metric to be used for the Principal Component Analysis. The Oblique Analysis is generalized to take into account this optimal metric. The interest of FAMIS approach is illustrated with MRI renal studies.