Foundations of factor analysis of medical image sequences: a unified approach and some practical implications

Factor Analysis of Medical Image Sequences (FAMIS) is presently conducted either in the function space or in the image space. A unified approach jointly using these two spaces is presented. First, the solution of a statistical model for scintigraphic image sequences leads to the use of correspondence analysis which is the optimal orthogonal decomposition of this data. Then, two symmetrical hypotheses concerning either the underlying fundamental functions or the underlying fundamental spatial distributions are derived. These hypotheses are merged in an original method to solve FAMIS physical model. Using this unified approach, a priori knowledge about functions and images can be jointly taken into account to improve the estimation of the underlying structures. Some practical applications of the method are illustrated on simulated data.