Session: P2F

been used to compare the commonly used polynomial regression (PR) filter with more advanced adaptive filtering techniques for clutter removal. More specifically, three adaptive filtering techniques were used in the comparison. The mean-frequency (MF) filter adapts to the tissue movement by down-mixing the temporal signal with the mean Doppler frequency. The varying-frequency (VF) filter adapts to the tissue movement by down-mixing the temporal signal with the varying phase increments for each time sample. Both filters subsequently apply a PR filter. The eigenvector regression (EV) filter spans the signal space with an eigenvector basis, where the high-energy clutter signal can be extracted by finding an appropriate basis for the clutter sub-space. The effectiveness of the clutter filters was quantified by comparing the signal power before and after filtering, from a region inside and outside a vessel. The procedure was repeated for increasing filter basis size. Results showed that all three adaptive filters outperformed the non-adaptive PR filter. A representative example from a region with accelerated muscle movement and a filter basis size of two, yielded a blood to clutter level of 11.2dB for the EV filter, 7.8dB for the VF filter, 2.5dB for the MF filter, and -6.1dB for PR filter. As the filter basis size was increased, the EV filter performance dropped to below that of all others in the study. A reason for this may be that the basis vectors spanning the blood subspace is then included and removed by the filter. In conclusion, all three adaptive filters ensured proper detection of blood, and a high reduction of clutter noise artefacts. For a basis size of one to three the VF and EV filter were equally the most effective. For a larger basis the VF filter prevailed. Also considering computational demands, the VF filter should be the overall filter of choice. The work presented was supported by the Medicine and Health program of the Norwegian Research Counsil.