A method for accurate in silico modeling of ultrasound transducer arrays

Abstract

This paper presents a new approach to improve the in silico modeling of ultrasound transducer arrays. While current simulation tools accurately predict the theoretical element spatio-temporal pressure response, transducers do not always behave as theorized. In practice, using the probe's physical dimensions and published specifications in silico, often results in unsatisfactory agreement between simulation and experiment. We describe a general optimization procedure used to maximize the correlation between the observed and simulated spatio-temporal response of a pulsed single element in a commercial ultrasound probe. A linear systems approach is employed to model element angular sensitivity, lens effects, and diffraction phenomena. A numerical deconvolution method is described to characterize the intrinsic electro-mechanical impulse response of the element. Once the response of the element and optimal element characteristics are known, prediction of the pressure response for arbitrary apertures and excitation signals is performed through direct convolution using available tools. We achieve a correlation of 0.846 between the experimental emitted waveform and simulated waveform when using the probe's physical specifications in silico. A far superior correlation of 0.988 is achieved when using the optimized in silico model. Electronic noise appears to be the main effect preventing the realization of higher correlation coefficients. More accurate in silico modeling will improve the evaluation and design of ultrasound transducers as well as aid in the development of sophisticated beamforming strategies.

Robust finite impulse response beamforming applied to medical ultrasound

“…Using DELFI and the single element model for the Ultrasonix L14-5/38mm probe developed in [16], a complete system model was simulated for a 64 element aperture focused at 3.0 cm in range assuming a speed of sound of 1540 m/s. ISRs were calculated with receive foci ranging from 0.5 cm to 5.0 cm spaced every 20 μm.…”

“…The FIR-QCLS weights computed in the previous section utilized the in silico model for the Ultrasonix Sonix RP scanner and the L14-5/38mm linear probe [16]. We now use these same weights to beamform some real, experimental data collected from a tissue mimicking phantom and a healthy human volunteer.…”

“…This analysis suggests that the “brute force” approach to determining every filter tap per output pixel via the QCLS algorithm may be excessive. The improvements in the experimental images indicate that the in silico model for the Sonix RP imaging system presented in [16] provided a sufficient degree of system characterization for the QCLS algorithm. Whether similar results could be achieved using a simpler, less sophisticated model or if further improvements could be achieved by incorporating nonlinearities, the elevation dimension, attenuation, etc.…”

“…A fully FIR beamformed image was not possible due to the large computational burden of constructing the filter taps for each output image pixel. In this paper, a complete in silico model [16] for the Ultrasonix Sonix RP scanner (Ultrasonix Medical Corp., Richmond, BC, Canada) using the L14-5/38 mm 128 element linear probe is used to generate 2,265 spatial impulse responses. These spatial responses are required to construct the FIR filters necessary to beamform an image from 0.5 cm to 5.0 cm in range.…”

On the Shaping of a Short Signal at the Output of the Receiving Piezoelectric Transducer in the Radiation-Reception System

“…Currently, there is a large number of theoretical works devoted to the analysis of the operating modes of piezoelectric transducers. Among them, we can note, for example, [ 23 , 24 , 25 , 26 , 27 , 28 , 29 ]. The number of works devoted to the study of piezoelectric transducers in the various fields of modern acoustics (biomedical research, hydroacoustics, flaw detection, etc.)…”