Speaker
Description
(2) Material and Methods
We first apply our FWI algorithm to a 3D simulated dataset generated with ultrasound transducers surrounding a realistic phantom of the breast. To efficiently invert the 3D dataset, we use a highly optimised finite differences forward modelling code that solves the isotropic acoustic wave equation. The dataset is generated with only a few hundred of transducers and the inversion is successfully sped up by considering less than a third of the total number of sources per iteration. We also implement our FWI code to a 2D in vivo dataset acquired with a single ring of 256 ultrasound transducers and with no information other than the recorded data, the total number of transducers and their nominal position.
(4) Discussion and Conclusion
We conclude that the presence of low frequencies allows for a much more efficient strategy when performing FWI of ultrasound data because it eliminates the need for a priori knowledge of the target and leads to higher quality recovered images. Thus, we encourage this community to use lower frequency transducers and to acquire data in three dimensions, which would allow more efficient and robust algorithms for 3D imaging providing images at a resolution enough for direct interpretation.
(3) Results
In contrast to reflection ultrasound tomography or time-of-flight tomography, where high-frequency transducers are necessary to obtain highly resolved models, we show that using low frequency ultrasound transducers − a few hundreds of kHz− and FWI is enough to obtain high-resolution recovered models. Despite not having low frequencies present in the 2D in vivo dataset, we are able to image a tumour and internal structures of the breast at higher resolution than conventional travel-time tomography. The inverted models also suggest that 3D data acquisition and inversion would be necessary to separate off-plane structures for a better interpretation and a more accurate diagnosis.
