Noninvasive estimation of maximum elastance from color-doppler m-mode echocardiograms using support vector machines

Peak systolic elastance (E_max) has been established as a quantitative measurement of left ventricular (LV) global systolic chamber function. However, a measurement of E_max is not possible in everyday clinical practice, due to the need of sophisticated catheterization procedures. Given that color-Doppler M-mode (CDMM) echocardiogram image represents the blood velocity, and given that fundamental hemodynamic magnitudes are related by complex physical laws, we hypothesize that E_max can be estimated noninvasively by adequate post-processing of CDMM. We propose to use support vector machines (SVM) for building a model based on CDMM velocity image. In an animal model (9 healthy pigs), several interventions were performed to obtain a range in E_max wider than basal values. CDMM images were acquired, together with E_max from catheters. Intraclass correlation coefficient for the combined independent test sets was 0.81 with the linear kernel and, surprisingly, lower (0.67) with the Gaussian kernel. In conclusion, the noninvasive estimation of E_max can be successfully addressed by using SVM regression on CDMM images