Deep data compression for approximate ultrasonic image formation

Pilikos, Georgios; Hörchens, Lars; Batenburg, Kees Joost; Leeuwen, Tristan van; Lucka, Felix

doi:10.1109/ius46767.2020.9251753

Cited by 9 publications

(3 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To alleviate this issue, a few contributions can be found on data compression. For instance, Pilikos et al [31] proposed an AE structure for data compression of ultrasonic array signals to later reconstruct them and create ultrasonic images. The method aims to enhance data transmission rates by enabling higher compression rates than classical approaches such as compressive sensing.…”

Section: Data Compressionmentioning

confidence: 99%

Deep learning in automated ultrasonic NDE -- developments, axioms and opportunities

Cantero-Chinchilla,

Wilcox,

Croxford

2021

Preprint

View full text Add to dashboard Cite

The analysis of ultrasonic NDE data has traditionally been addressed by a trained operator manually interpreting data with the support of rudimentary automation tools. Recently, many demonstrations of deep learning (DL) techniques that address individual NDE tasks (data acquisition, data pre-processing, defect detection, and defect characterisation) have started to emerge in the research community. These methods have the potential to offer high flexibility, efficiency, and accuracy subject to the availability of sufficient training data; moreover, they enable the automation of complex processes that span one or more NDE steps (e.g. detection, characterisation, and sizing). There is, however, a lack of consensus on the direction and requirements that these new methods should follow. These elements are critical to help achieve full artificial intelligence driven automation of ultrasonic NDE so that the research community, industry, and regulatory bodies embrace them. This paper reviews the state-of-the-art of autonomous ultrasonic NDE enabled by DL methodologies. The review is organised by the NDE tasks that are addressed by means of DL approaches. Key remaining challenges for each task are noted. Basic axiomatic principles for DL methods in NDE are identified based on the literature review, relevant international regulations, and current industrial needs. By placing DL methods in the context of general NDE automation levels, this paper aims to provide a roadmap for future research and development in the area.

show abstract

Section: Data Compressionmentioning

confidence: 99%

Deep learning in automated ultrasonic NDE -- developments, axioms and opportunities

Cantero-Chinchilla,

Wilcox,

Croxford

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Alternative deep learning techniques have been proposed that combine the great expressive power of data-driven neural networks with traditional image formation techniques. In particular, deep learning and the DAS algorithm have been combined in an end-to-end deep learning framework, obtaining superior results compared to purely datadriven models for both imaging and segmentation [11] [12].…”

Section: Introductionmentioning

confidence: 99%

Single Plane-Wave Imaging using Physics-Based Deep Learning

Pilikos,

de Korte,

van Leeuwen

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In plane-wave imaging, multiple unfocused ultrasound waves are transmitted into a medium of interest from different angles and an image is formed from the recorded reflections. The number of plane waves used leads to a tradeoff between frame-rate and image quality, with single-planewave (SPW) imaging being the fastest possible modality with the worst image quality. Recently, deep learning methods have been proposed to improve ultrasound imaging. One approach is to use image-to-image networks that work on the formed image and another is to directly learn a mapping from data to an image. Both approaches utilize purely data-driven models and require deep, expressive network architectures, combined with large numbers of training samples to obtain good results.Here, we propose a data-to-image architecture that incorporates a wave-physics-based image formation algorithm in-between deep convolutional neural networks. To achieve this, we implement the Fourier (FK) migration method as network layers and train the whole network end-to-end. We compare our proposed datato-image network with an image-to-image network in simulated data experiments, mimicking a medical ultrasound application. Experiments show that it is possible to obtain high-quality SPW images, almost similar to an image formed using 75 plane waves over an angular range of ±16 • . This illustrates the great potential of combining deep neural networks with physics-based image formation algorithms for SPW imaging.

show abstract

“…Digital time-domain multiplexing (TDM), on the other hand, has been shown to benefit from better tolerance to cross-talk, interference and noise [15]. The availability of digital receive signals in the catheter moreover opens the possibility for future co-integration with emerging image processing such as data reduction with machine-learned compression [24], [25] or adaptive beamforming [26]. A major benefit of multiplexing lies in the compatibility with subarray beamforming, as has been shown in digital beamforming of element-level signals [27]- [29] and analog beamforming with subsequent digitization and TDM [15], [30], [31].…”

mentioning

confidence: 99%

A Pitch-Matched Transceiver ASIC With Shared Hybrid Beamforming ADC for High-Frame-Rate 3-D Intracardiac Echocardiography

Hopf

Ossenkoppele

Soozande

et al. 2022

IEEE J. Solid-State Circuits

View full text Add to dashboard Cite

In this paper, an application-specific integrated circuit (ASIC) for 3D, high-frame-rate ultrasound imaging probes is presented. The design is the first to combine element-level, high-voltage (HV) transmitters and analog frontends, subarray beamforming, and in-probe digitization in a scalable fashion for catheter-based probes. The integration challenge is met by a hybrid analog-to-digital converter (ADC), combining an efficient charge-sharing successive-approximation-register (SAR) first stage and a compact single-slope (SS) second stage. Application in large ultrasound imaging arrays is facilitated by directly interfacing the ADC with a charge-domain subarray beamformer, locally calibrating inter-stage gain errors and generating the SAR reference using a power-efficient local reference generator. Additional hardware-sharing between neighboring channels ultimately leads to the lowest reported area and power consumption across miniature ultrasound probe ADCs. A pitchmatched design is further enabled by an efficient split between the core circuitry and a periphery block, the latter including a datalink performing clock-data-recovery (CDR) and timedivision multiplexing (TDM), which leads to a 12-fold total channel-count reduction. A prototype of 8×9 elements was fabricated in TSMC 0.18-µm HV BCD technology and a 2D PZT transducer matrix with a pitch of 160 µm and a center frequency of 6 MHz was manufactured on the chip. The imaging device operates at up to 1000 volumes/s, generates 65-V transmit pulses and has a receive power consumption of only 1.23 mW/element. The functionality has been demonstrated electrically as well as in acoustic and imaging experiments.

show abstract

Deep data compression for approximate ultrasonic image formation

Cited by 9 publications

References 14 publications

Deep learning in automated ultrasonic NDE -- developments, axioms and opportunities

Deep learning in automated ultrasonic NDE -- developments, axioms and opportunities

Single Plane-Wave Imaging using Physics-Based Deep Learning

A Pitch-Matched Transceiver ASIC With Shared Hybrid Beamforming ADC for High-Frame-Rate 3-D Intracardiac Echocardiography

Contact Info

Product

Resources

About