Normality Index of Ventricular Contraction Based on a Statistical Model from FADS
Figure 5
Schematic representation of the proposed methodology. (a). The normal contraction pattern was modeled as a linear combination of Gaussian density functions. The relative weights , means () and covariance matrices () of the Gaussian mixture were estimated maximizing the mixture model likelihood of the probability density functions of the 3 most significant factors (F1, F2, and F3) computed from the Factor Analysis of Dynamic Structures (FADS) and following a standard bootstrap resampling procedure with a set of 23 control subjects. (b) Considering that the likelihood estimated on a data sample represents the probability that those observations are well described by the assumed model with , , and parameters, we propose indices based on this probability. A normality index () for a group of normal subjects (not considered for the mixture parameters estimation) was measured, and statistically compared with the indices obtained for LBBB and DCM populations.
