Super-Resolution Ultrasound Localization Microscopy Through Deep Learning

Abstract

DOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the "Taverne" license above, please follow below link for the End User Agreement:

Detection and Localization of Ultrasound Scatterers Using Convolutional Neural Networks

“…To compare the proposed method with Deep-ULM, the recall and localization errors were recalculated following the method which van Sloun et al used to generate the results in the supplementary Fig. 1 in [24]. The threshold value for determining positive detection was λ/7 and Euclidean distances between the true and estimated scatterers were calculated.…”

“…4b) was proposed to solve the imbalance problem of the binary confidence maps. Applying 2-D Gaussian filtering to sparse labels can improve training stability and guide CNNs to correct solutions [21], [24], [39]. But simply applying 2-D Gaussian filtering is problematic because the scatterer positions cannot be recovered in the confidence maps when the scatterers are closely spaced, as shown in Fig.…”

“…Similar attempts also exist in SRUS. Van Sloun et al [24] proposed Deep-ULM that outputs high-resolution images where the pixel values correspond to scattering intensities, given image patches of contrast-enhanced ultrasound (CEUS) acquisitions. This is similar to our approach in the sense that it handles high-density scatterer detection using CNNs, but Deep-ULM takes beamformed signals as input, whereas the proposed method only uses RF channel data without beamforming.…”

Current Development and Applications of Super-Resolution Ultrasound Imaging

“…An increased concentration of microbubbles would reduce overall data acquisition time. The Deep-ULM, inspired by the deep learning network for super-resolution stochastic optical-resolution microscopy (Deep-STORM), adopts a network based on the fully convolutional U-net, performing the nonlinear end-to-end mapping between low-resolution input frames to high-resolution outputs, as shown in Figure 3 [ 39 , 47 , 48 , 49 ]. For the synthetic training dataset, randomly located microbubble positions were generated first.…”

Noncontact super-resolution guided wave array imaging of subwavelength defects using a multiscale deep learning approach

“…Deep learning has been actively applied to improve the imaging resolution in various imagingrelated fields, including fluorescence microscopy, 25 Xray tomography, 26 photoacoustic tomography, 27 and ultrasound localization microscopy. 28 Recently, it has also been employed in NDE and SHM applications. [29][30][31] Despite recent active implementations of deep learning in various imaging-related fields, to the authors' best knowledge, the deep learning approach for superresolution guided wave array imaging has not been studied yet.…”