Deep-Learning-Driven Full-Waveform Inversion for Ultrasound Breast Imaging

Abstract

Ultrasound breast imaging is a promising alternative to conventional mammography because it does not expose women to harmful ionising radiation and it can successfully image dense breast tissue. However, conventional ultrasound imaging only provides morphological information with limited diagnostic value. Ultrasound computed tomography (USCT) uses energy in both transmission and reflection when imaging the breast to provide more diagnostically relevant quantitative tissue properties, but it is often based on time-of-flight tomography or similar ray approximations of the wave equation, resulting in reconstructed images with low resolution. Full-waveform inversion (FWI) is based on a more accurate approximation of wave-propagation phenomena and can consequently produce very high resolution images using frequencies below 1 megahertz. These low frequencies, however, are not available in most USCT acquisition systems, as they use transducers with central frequencies well above those required in FWI. To circumvent this problem, we designed, trained, and implemented a two-dimensional convolutional neural network to artificially generate missing low frequencies in USCT data. Our results show that FWI reconstructions using experiment data after the application of the proposed method successfully converged, showing good agreement with X-ray CT and reflection ultrasound-tomography images.

The Emerging Role of Circulating Tumor DNA in the Management of Breast Cancer

“…This inaccuracy becomes more apparent with computer-aided detection (augmented with artificial intelligence), which has been shown to improve the overall accuracy and diagnostic specificity of mammography compared to non-computeraided detection [41]. In addition, applications of artificial intelligence have been applied to ultrasound (US) imaging in order to better detect breast cancer [42,43]. Novel techniques in high-resolution, low frequency US and neural networks to analyze densely pixelated images have been reported to correlate with radiologist diagnostic interpretations of surveillance imaging for patients with breast cancer.…”

Dense speed-of-sound shift imaging for ultrasonic thermometry

“…Instead, we propose unifying echo shift information from several plane waves to predict the change in speed-of-sound using an inverse problem method regularized with a spatial smoothness prior, which can similarly be calibrated to yield the underlying temperature shift. Regularized inverse problem approaches have shown promising results in ultrasound applications related to sound speed (Stähli et al 2020, Robins et al 2021, in addition to additional methods based on the beam geometry (Brevett et al 2022), passive reflectors (Sanabria et al 2019) or diverging waves (Rau et al 2021).…”

Simulation-to-real generalization for deep-learning-based refraction-corrected ultrasound tomography image reconstruction

“…It is believed that FWI has the potential for high resolution image reconstruction (Lucka et al 2021). However, to get a high fidelity reconstruction, FWI needs good initialization and low frequency information(a few hundreds of kHz) (Agudo et al 2018) to avoid cycle skipping (Robins et al 2021, Boehm et al 2022. This low frequency information is often unavailable in conventional ultrasound tomography machines.…”

“…In recent years, there has been a growing interest in deep learning methods to allow for real-time USCT image reconstruction, which can be divided into two categories: hybrid approach and fully-learned approach. The hybrid approach aims at integrating deep learning methods with traditional iterative reconstruction methods to achieve faster reconstruction (Poudel et al 2019, Robins et al 2021, Stanziola et al 2021, Fan et al 2022. On the other hand, the fully learned approach tries to reconstruct images via end-to-end learning with deep learning methods from measurement data (Prasad andAlmekkawy 2020, Zhao et al 2020).…”