SPARSE 2 D ARRAYS FOR REAL-TIME 3 D ULTRASOUND

Existing 3D ultrasound systems are based on mechanically moving 1D arrays for data collection and post-processing of data to achieve 3D images. To be able to both collect and process 3D data in real-time, a scaling of the ultra-sound system from 1D to 2D arrays is necessary. A typical 2D-system uses 1D arrays with about 100 receive and transmit channels. Scaling of this system to get a 3D-system with as good performance as the 2D-system implies a squaring of the number of channels, i.e. a 10.000 channel system. To reduce cost and complexity of such a system, removal of array elements or equally channels is possible. Arrays with removed elements are known as sparse arrays. At the University of Oslo, there are two ongoing projects which aim to find optimal 2D sparse layouts through optimization and simulation. The goal is to minimize the number of channels without compromising image quality. To verify this work and to critically test system performances , an extensive evaluation program is to be carried out. This includes collection of experimental data from a 3D prototype array by using a state-of-the-art digital 2D ultrasound scanner and massive post processing on a cluster of workstations. To simplify future integration of electronics into the probe, the sparse transmit and receive layouts should be chosen to be non-overlapping. This means that some elements should be dedicated to transmit while others should be used to receive. To increase system performance, future 2D-arrays should possibly include pre-amplifiers directly connected to the receive elements. The problem of finding the optimal sparse layout is enormous. If the array consists of 50×50 elements, and 80% of the elements should be removed, then there are more than 10 541 ways to choose the 500 out of the 2500 elements. A 50 × 50 element array usually has a quadratic footprint. If one limits the number of elements by introducing a circular footprint, and takes into account that both transmit and receive layout have to be chosen, and these should be non-overlapping, the number of combinations equals 1964 500 × 1464 500 ≈ 10 888. As a comparison, there are believed to be 10 80 electrons in the universe. The layouts should have optimal pulse-echo performance , i.e. the pulse-echo radiation pattern should have as low sidelobe level as possible for a specified main lobe width for all angles and depths of interest. To compute the …