Session: 14 - 01 Ultrasonic Arrays I

Paper Number: 108664

108664 - Ultrasound Imaging by Accelerated Data Inversion With Arbitrary Waveform Emission. 

In ultrasonic non-destructive testing (NDT), the performances of imaging devices are often evaluated as a compromise between the speed of the acquired ultrasonic data and the quality (resolution, contrast) of the reconstructed images.
Standard beamforming techniques primarily relying on Delay-And-Sum (DAS), such as the total focusing method (TFM), are often used in real-time applications. However, the estimated images suffer from limited spatial resolution and contrast.
Recently, post-processing imaging techniques based on inverse problems have been proposed to improve the quality of the images, but with an often high computation time due to a large amount of data.
In this work, we aim to accelerate the reconstruction process by adopting the screening principle, which exploits properties related to the sparsity of the solution without compromising the quality of the image. The ultimate goal is to apply imaging based on inverse problems in real-time applications. In addition, we propose a linear forward model that allows different acquisition schemes, such as plane wave imaging (PWI), full matrix capture (FMC), or simultaneous transmission. This model considers the transducer response and specific excitation signals (pulse, coded signals with Chirp, random sequences, etc.).
The reconstruction algorithm uses the acquired data and the developed forward model to estimate the reflectivity map. We propose to use the least square approach with L1 regularization to enhance sparsity, solved by the optimization algorithm FISTA. In addition, we propose using screening techniques to accelerate the reconstruction process by predicting and discarding elements that do not contribute to the problem solution.
We first validate the performance of our algorithm from synthetic data with the developed forward model. We consider a probe array with 64 elements in contact with a homogeneous medium containing close reflectors separated by a wavelength over two. We perform several tests with different excitation signals and acquisition models while varying the SNR and the image size. We also test the proposed algorithm on experimental FMC data acquired using a 128-channel Pioneer TPAC system. The sample under test is an aluminum block with two close side-drilled holes (SDH) of 1 mm diameter separated by 1 mm. The probe has 128 elements with a center frequency of 3 MHz and a pitch of 0.8 mm. The Rayleigh criterion is 1.86 mm, about twice the distance between the defects.
The proposed algorithm significantly improves the quality of the images reconstructed from synthetic and experimental data compared to the standard TFM method. Moreover, adding the screening rule to the optimization algorithm reduces the computation time by about a factor of two while preserving the quality of the reconstructed image.

Presenting Author: Ralph Abi Rizk TPAC

Presenting Author Biography: I received my Ph.D. in image and signal processing from the L2S laboratory in Nantes, France, in 2021. I am currently a Postdoctoral researcher with LS2N laboratory, Centrale-Nantes, Nantes, France, and TPAC company, Nantes, France.
My research interests include inverse problem approaches, ultrasonic image processing for non-destructive testing, and hyperspectral image reconstruction.