Energy and feature-based strategies for estimating the contrast agent concentration in ultrasound medical imaging

The estimation of the contrast agent concentration is important for the introduction of effective quantification in ultrasound medical diagnoses. This paper focuses on the comparison of three methods for estimating the number of microbubbles in a small volume. An energy-based linear regression approach and a nonlinear regression approach are applied to estimate the microbubble concentration working from the signals backscattered by a contrast agent population. The nonlinear regression approach, based on the support vector machine, is employed assuming either only one feature, the signal energy, or a set of significant features. Synthetic data are used for the training phase and the test of the methods. The results demonstrate that, although the energy-based methods allows to achieve good results in terms of mean estimated values, the adoption of a vector of features significantly reduces the mean square error. The nonlinear regression approach with multidimensional feature vector allows to improve the measurement precision of the bubble concentration.