Adaptive ultrasound clutter rejection through spatial eigenvector filtering

Off-axis clutter is a significant cause of image degradation in ultrasound. We propose an adaptive approach for clutter reduction based on the principle that off-axis signals show as relatively high spatial frequencies in the per-channel data. For each pixel, an M* Nper-channel data matrix is extracte from the aligned data received from different observations of that pixel (N = number of channels, M = number of observations). Singular Value Decomposition (SVD) of that matrix yields the eigenvectors, which correspond to the most important "modes" of the per-channel data. The mean spatial frequency of each eigenvector is estimated, and the eigenvectors with relatively high mean spatial frequencies, assumed to correspond to off-axis signals, are attenuated prior to SVD reconstruction. The filtered per-channel data thus obtained are then summed to yield a pixel value estimate with attenuated contribution from the off-axis signals. The concept is proven on simulation, phantom, and in vivo cardiac data.