Abstract
Observing the correlation between the information in neighboring resolution cells, we propose a methodology to allow better analysis of Doppler ultrasound signals. A blind source separation problem is formulated to discern the different signal components for correct interpretation of the data using independent component analysis (ICA). The Doppler signal is modeled as the summation of the true velocity signal, baseline fluctuation, and random noise. The baseline fluctuation component can be considered as a deterministic yet unknown signal. A simple adaptive denoising technique is applied to reduce the effective dimension of the noise subspace. Then, given a region of interest, the temporal signals corresponding to all pixels within this region undergo the ICA iteration to compute a set of independent signals that most represent the actual components present within the data. Subsequently, a comparison of the temporal variations of these signals allows the user identify the components of the signals that correspond to baseline variation or random noise manually or using semi-automated techniques. The new technique shows large potential to alleviate some of the limitations in this demanding imaging mode as well as to make the interpretation of the results more robust.