Row-Column Beamformer for Fast Volumetric Imaging

A row-column beamformation algorithm is presented, which yields the output from the conventional delay-and-sum algorithm while reducing the number of operations by order of magnitude. The proposed method uses that in a row-column synthetic aperture sequence, the low-resolution volumes (LRV) have approximately constant image values along the elevation axis. It is, thus, possible to reconstruct the entire LRV from a single cross-section by modeling the positions with constant image values. As such, the proposed method consists of two stages. The first stage beamforms a low-resolution image per emission using the conventional approach. The second stage reconstructs the LRVs with one interpolation per voxel. Lastly, a high-resolution volume (HRV) is obtained by summing the LRVs across all emissions. The proposed algorithm was evaluated on measured data acquired using a 6 MHz 128+128 Vermon row-column probe and a Verasonics Vantage system. The proposed method beamformed a $100 \times 100 \times 200$ HRV consisting of 48 LRVs at a volume rate of 38 Hz. This was 9.23 times faster than a published GPU implementation of the conventional approach and real-time volumetric beamformation was achieved with a pulse repetition frequency of up to 1805 Hz. The output from the conventional and proposed beamformer was visually indistinguishable, and their point spread function's width heccurate interpolation requiright at -6 dB and -20 dB deviated less than 0.5%. This demonstrates that the number of operations of the conventional row-column beamformer can be significantly reduced with a negligible impact on image quality.