Beamforming and Speckle Reduction Using Neural Networks

Hyun, Dongwoon; Brickson, Leandra; Looby, Kevin; Dahl, Jeremy J.

doi:10.1109/tuffc.2019.2903795

Cited by 150 publications

(75 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The success of the presented results has implications for providing multiple (i.e., more than two) DNN outputs from a single network input. For example, in addition to beamforming and segmentation, deep learning ultrasound image formation tasks have also been proposed for sound speed estimation [60], speckle reduction [43], reverberation noise suppression [61], and minimum-variance directionless response beamforming [62], as well as to create ultrasound elastography images [63], CT-like ultrasound images [64], B-mode images from echogenecity maps [65], and ultrasound images from 3D spatial locations [66]. We envisage the future use of parallel networks that output any number of these or other mappings to provide a one-step approach to obtain multimodal information, each originating from a singular input of raw ultrasound data.…”

Section: Discussionmentioning

confidence: 99%

Deep Learning to Obtain Simultaneous Image and Segmentation Outputs From a Single Input of Raw Ultrasound Channel Data

Nair

Washington

Tran

et al. 2020

IEEE Trans. Ultrason., Ferroelect., Freq. Contr.

View full text Add to dashboard Cite

Single plane wave transmissions are promising for automated imaging tasks requiring high ultrasound frame rates over an extended field of view. However, a single plane wave insonification typically produces suboptimal image quality. To address this limitation, we are exploring the use of deep neural networks (DNNs) as an alternative to delay-and-sum beamforming. The objectives of this work are to obtain information directly from raw channel data and to simultaneously generate both a segmentation map for automated ultrasound tasks and a corresponding ultrasound B-mode image for interpretable supervision of the automation. We focus on visualizing and segmenting anechoic targets surrounded by tissue and ignoring or de-emphasizing less important surrounding structures. DNNs trained with Field II simulations were tested with simulated, experimental phantom, and in vivo datasets that were not included during training. With unfocused input channel data (i.e., prior to the application of receive time delays), simulated, experimental phantom, and in vivo test datasets achieved mean ± standard deviation Dice similarity coefficients of 0.92 ± 0.13, 0.92±0.03, and 0.77±0.07, respectively, and generalized contrastto-noise ratios (gCNR) of 0.95±0.08, 0.93±0.08, and 0.75±0.14, respectively. With subaperture beamformed channel data and a modification to the input layer of the DNN architecture to accept these data, the fidelity of image reconstruction increased (e.g., mean gCNR of multiple acquisitions of two in vivo breast cysts ranged 0.89-0.96), but DNN display frame rates were reduced from 395 Hz to 287 Hz. Overall, the DNNs successfully translated feature representations learned from simulated data to phantom and in vivo data, which is promising for this novel approach to simultaneous ultrasound image formation and segmentation.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep Learning to Obtain Simultaneous Image and Segmentation Outputs From a Single Input of Raw Ultrasound Channel Data

Nair

Washington

Tran

et al. 2020

IEEE Trans. Ultrason., Ferroelect., Freq. Contr.

View full text Add to dashboard Cite

show abstract

“…Beyond beamforming for suppression of off-axis scattering, the authors in [44] propose deep convolutional neural networks for joint beamforming and speckle reduction. Rather than applying the latter as a post-processing technique, it is embedded in the beamforming process itself, permitting exploitation of both channel and phase information that is otherwise irreversibly lost.…”

Section: Deep Learning For (Front-end) Ultrasound Processingmentioning

confidence: 99%

Deep Learning in Ultrasound Imaging

2020

View full text Add to dashboard Cite

Deep learning is taking an ever more prominent role in medical imaging. This paper discusses applications of this powerful approach in ultrasound imaging systems along with domain-specific opportunities and challenges.ABSTRACT | We consider deep learning strategies in ultrasound systems, from the front-end to advanced applications. Our goal is to provide the reader with a broad understanding of the possible impact of deep learning methodologies on many aspects of ultrasound imaging. In particular, we discuss methods that lie at the interface of signal acquisition and machine learning, exploiting both data structure (e.g. sparsity in some domain) and data dimensionality (big data) already at the raw radio-frequency channel stage.As some examples, we outline efficient and effective deep learning solutions for adaptive beamforming and adaptive spectral Doppler through artificial agents, learn compressive encodings for color Doppler, and provide a framework for structured signal recovery by learning fast approximations of iterative minimization problems, with applications to clutter suppression and super-resolution ultrasound. These emerging technologies may have considerable impact on ultrasound imaging, showing promise across key components in the receive processing chain.

show abstract

“…In this section, the test dataset will be applied with different techniques of transformations to evaluate the potential performance degradation. Gaussian noise [68], salt & pepper noise [69], speckle noise [70], and image rotation [71] are chosen to artificially transform the test images iteratively with different settings to study the impacts. Fig.…”

Section: Sensitivity Analysismentioning

confidence: 99%

Quantized Deep Residual Convolutional Neural Network for Image-Based Dietary Assessment

2020

View full text Add to dashboard Cite

Vegetable intake is an essential element to maintain a healthy body of a human. However, research shows most people do not consume an adequate intake of vegetables per day. An ameliorate dietary assessment for vegetable intake is needed to increase awareness and assist users to improve their vegetable consumption. In this paper, we proposed a novel Quantized Deep Residual Convolutional Neural Network (DRCNN) model to ameliorate the fundamental task of dietary assessment. The proposed deep learning strategy integrating a deep residual architecture and a deep learning model, i.e. convolutional neural network, to apply for image-based dietary assessment. The proposed DRCNN model is then deeply fused with posttraining quantization techniques to quantize the network weights of the proposed DRCNN model into lowbit fixed-point representations. Extensive experiments have been done to evaluate the proposed model. Results show that the proposed Quantized DRCNN model outperformed the state-of-the-arts, which include the conventional CNN models and also the machine learning models. Those experiments also indicate the effectiveness of the proposed Quantized DRCNN model. Finally, we further evaluate our model with crossdataset validation to verify the generalization of the proposed model. The experimental result proves that the proposed Quantized DRCNN model is general enough to predict unseen cases and it works well on a wide range of food images. INDEX TERMS Convolutional neural network, deep learning, deep residual network, image-based dietary assessment, machine learning, quantized model.

show abstract

Beamforming and Speckle Reduction Using Neural Networks

Cited by 150 publications

References 35 publications

Deep Learning to Obtain Simultaneous Image and Segmentation Outputs From a Single Input of Raw Ultrasound Channel Data

Deep Learning to Obtain Simultaneous Image and Segmentation Outputs From a Single Input of Raw Ultrasound Channel Data

Deep Learning in Ultrasound Imaging

Quantized Deep Residual Convolutional Neural Network for Image-Based Dietary Assessment

Contact Info

Product

Resources

About