ABSTRACT
A new technique for studying quantitatively the spatial geometry of the left ventricle (LV) is presented here. The technique utilizes a helical-shape descriptor which transforms the three-dimensional geometrical information into a unidimensional vector mode, R(ξ). Applying an anatomical alignment and a normalization procedure to the R(ξ) vector yields a dimensionless signal, defined as the geometrical cardiogram (GKG). The GKG vector contains the basic features of the LV spatial geometry, is independent of LV size, and thus allows a quantitative comparison between similar and dissimilar shapes. Furthermore, applying the Karhunen-Loeve transform (KLT) to a clinically defined “characteristic” group of LVs yields a typical, “canonical” three-dimensional shape of the group. The canonical shape technique may serve to compare other GKGs to normal hearts and for shape sorting of the instantaneous LV geometry according to various pathologies. Utilizing the Fourier transform for analysis of the individual GKG at end-diastole (ED) and end-systole (ES) in the frequency domain provides information which characterizes the LV global contraction. A similar analysis in the spatial domain provides information about the local function of the myocardium, thus enabling detection and definition of akinetic sites on the LV wall.
