|M.Sc Student||Soceanu Omri|
|Subject||Thinning Transducer Arrays for Ultrasound Imaging using|
|Department||Department of Electrical Engineering||Supervisors||Professor Moshe Porat|
|Dr. Zvi Friedman|
Ultrasound systems are generally comprised of back-end and front-end processing modules. While the transmitting and receiving elements are located in the front-end module, the beamforming can be performed in either module. Each approach has its advantages and disadvantages. Performing the beamforming in the back-end module allows for better performance. Utilizing all the receiving elements’ signal information during the signal processing stage, prior to the beamforming process, allows the ultrasound designer to employ a range of quality enhancing techniques. These include various algorithms such as Adaptive Beamforming, Phase Coherence Factor Multiplication, Retrospective Dynamic Transmit Focusing, Short Lag Spatial Coherence and many others. However, using these algorithms comes with a price, power consumption. Though in some cases this challenge is solvable (e.g. large static ultrasound systems), this is not the case for battery powered portable devices.
The challenge of producing high-resolution and high frame-rate ultrasound video streams while maintaining a high-quality picture through signal processing algorithms, mainly clutter rejection, requires multi-individual channel processing. Whether these algorithms are implemented at the front-end or back-end modules, they drive up power consumption in the front-end module. Power-demanding analog to digital converters, and either a high-power front-end CPU or high-power intra-module communication lines are used to support the increase in data throughput. Ultrasound devices leverage a sensor array to improve signal quality through beamforming. Since data throughput in these cases is directly linked to the number of ultrasonic elements in the receiver array, thinning their number without compromising the image quality has been an obvious yet elusive solution. Radio telescope designers have faced similar challenges and have used multiplicative beamforming for the construction of thinned antenna arrays. We have translated this concept to medical ultrasound imaging. Accordingly, through this nonlinear multiplicative processing, we have approximated the beam profile of an evenly spaced ultrasonic array by multiplying and then filtering two sub-arrays of elements: one consisting of a short-filled array of elements and a centered-aligned thinned array. This concept has never been implemented in ultrasonic imaging before because any reduction in the number of elements reduces both signal to noise ratio and contrast resolution, and the high dynamic range required could not be achieved without access to and processing of individual channel data. In this work, however, we took advantage of the very large reduction in the number of channels to implement computationally intensive signal processing algorithms that would have been unfeasible in power restricted systems.
Leveraging the smaller number of signal processing channels, we have designed and implemented an ultrasound system composed of a series of signal processing algorithmic modules that aim to improve image quality beyond what is possible in a full-array ultrasound system. We further addressed the high dynamic range issue by separating out the strong reflectors and processing them in a separate channel.
Our conclusion is that applying the required modifications, multiplicative beamforming can be successfully translated to ultrasound imaging allowing a significant reduction in the number of individual signal processing channels, without practically compromising performances.