Signal Processing in Two-Dimensional Arrays for 3-D imaging

Sverre Holm, Department of Informatics, University of Oslo


Imaging in three dimensions with a quality and an update rate equal to today's two dimensional imaging is the objective of on-going real-time 3-D work. This requires 2-D arrays for the data acquisition, because their greater beam agility is a prerequisite for maximizing the volume update rate.

An area of concern is the cost of the front-end of the ultrasound system. Therefore thinned arrays have been widely studied in order to get the channel count down. We have used algorithmic optimization of the layout using linear programming, genetic optimization, and simulated annealing algorithms. For the objective of minimizing the maximum sidelobe level with a constraint on the mainlobe width, we have worked on finding lower limits on sidelobe level. The result is that when the same elements are used for reception and transmission, the two-way levels will not be acceptable for ultrasound imaging, unless the element count after sparsing is several thousands.

A better approach is to allow different elements for the receiver and the transmitter. Sparse periodic arrays are based on this principle. If for instance every third element is used for the receiver and every second for the transmitter, the resulting grating lobes will partly cancel. There will still be sidelobes or grating lobes that will affect image quality in a negative way, but the result is an improvement for the same element count. Due to the requirement for different periodicities, it is inherent in this method that some elements have to be shared between the transmitter and the receiver.

Overlapping elements can be avoided if a combination of the two methods is used, e.g. a periodic array for transmission and an algorithmically optimized array for the receiver. The condition to avoid overlap can easily be included in an optimization. This kind of array can give image quality close to that of the periodic array. Equally important is that it is better suited for integration of electronics in the transducer. Dedicated receive or transmit circuitry can be integrated with the elements assigned to either reception or transmission. We are in the process of testing all these sparsing methods on a 50 x 50 test transducer.

It should, however, be kept in mind that the penalty of sparsing is image quality degradation, so eventually it is desirable to use full 2-D array. Emergence of a new technology such as micromachined silicon transducers may make this feasible and change the rules of the game completely. If such transducers turn out to be competitive with piezo-ceramic transducers in acoustic performance and in cost, they may allow simple integration of electronic circuitry, and probe manufacturing using processes from the semiconductor industry.

Another fundamental problem is that of getting around the limited velocity of sound that is the cause of the volume rate problem in 3-D ultrasound. The problem is how to get parallelism without image quality degradation. A solution to this problem may be based on a combination of all the methods proposed: Parallel receive beams, synthetic aperture imaging, multiple transmit beams and limited diffraction X-waves.