Abstract
Singular value decomposition (SVD)-based clutter filters have shown its outstanding ability in clutter rejection for ultrafast plane-wave Doppler imaging. However, automatic decision on order selection still remains a big problem in vivo and the computational complexity of the SVD-based clutter filter is also high for clinical translation. In this study, we propose a novel convolutional autoencoder based clutter filtering technique for power Doppler, which retains the characteristic principle component analysis (PCA) done by the SVD filtering while being with no need for order selection and potentially lower computational load. The autoencoder is a counterpart for PCA in deep learning field. The simulation results show that the proposed convolutional autoencoder clutter filtering technique can robustly reject tissue clutter whose spectrum overlaps that of the blood signal, which is impossible to accomplish with conventional highpass filtering, and outperforms the SVD filtering. Our simulation results show that the contrast-to-noise ratio (i.e., blood-to-clutter ratio) of our proposed method is better than those of the SVD-based filter and the highpass filter by 23.97 dB and 64.17 dB, respectively.