Publications Search Results
Now showing 1 - 10 of 12
PublicationDeep Learning Based Angular Compounding for Accelerated Plane Wave Ultrasound Imaging( 2021)
;Strohm, H. ;Rothlübbers, S. ;Jenne, J.Günther, M.The quality of ultrasound plane wave imaging benefits from compounding multiple angle acquisitions to reconstruct an image. However, the acquisition of additional data lowers the frame rate and - in presence of motion - the data integrity. This work presents an approach to reconstruct high-quality images from a reduced set of angles making use of artificial deep neural networks (DNNs). Unlike existing approaches that utilize DNNs for transforming beamformed data into image data directly, the presented DNN is trained to produce per-pixel angular weighting factors within an existing reconstruction pipeline.
PublicationDeep Learning for Ultrasound Image Formation: CUBDL Evaluation Framework and Open Datasets( 2021)
;Hyun, D. ;Wiacek, A. ;Goudarzi, S. ;Rothlübbers, S. ;Asif, A. ;Eickel, K. ;Eldar, Y.C. ;Huang, J. ;Mischi, M. ;Rivaz, H. ;Sinden, D. ;Sloun, R.J. van ;Strohm, H.Lediju Bell, M.A.Deep learning for ultrasound image formation is rapidly garnering research support and attention, quickly rising as the latest frontier in ultrasound image formation, with much promise to balance both image quality and display speed. Despite this promise, one challenge with identifying optimal solutions is the absence of unified evaluation methods and datasets that are not specific to a single research group. This paper introduces the largest known international database of ultrasound channel data and describes associated evaluation methods that were initially developed for the Challenge on Ultrasound Beamforming with Deep Learning (CUBDL), which was offered as a component of the 2020 IEEE International Ultrasonics Symposium. We summarize the challenge results and present qualitative and quantitative assessments using both the initially closed CUBDL evaluation test dataset (which was crowd-sourced from multiple groups around the world) and additional in vivo breast ultrasound data contributed after the challenge was completed. As an example quantitative assessment, single plane wave images from the CUBDL Task 1 dataset produced a mean generalized contrast-to-noise ratio (gCNR) of 0.67 and a mean lateral resolution of 0.42 mm when formed with delay-and-sum beamforming, compared to a mean gCNR as high as 0.81 and a mean lateral resolution as low as 0.32 mm when formed with networks submitted by the challenge winners. We also describe contributed CUBDL data that may be used for training of future networks. The compiled database includes a total of 576 image acquisition sequences. We additionally introduce a neural network-based global sound speed estimator implementation that was necessary to fairly evaluate results obtained with this international database.
PublicationBeyond classical ultrasound contrast via deep neural networks( 2020)
;Strohm, H. ;Rothlübbers, S. ;Eickel, K.Günther, M.Classical ultrasound reconstruction applies model driven approaches to obtain ultrasound images from ultrasound raw data. With the emergence of Deep Learning however data driven approaches become feasible and can be explored. These can be used to take shortcuts in the reconstruction, directly learning the relationship between raw data and image data. Even more, entirely new target contrasts can be pursued. In this work we present an approach to train a neural network to reconstruct image data of a classical ultrasound and a novel MR-like contrast from the same ultrasound raw data.
PublicationImproving image quality of single plane wave ultrasound via deep learning based channel compounding( 2020)
;Rothlübbers, S. ;Strohm, H. ;Eickel, K. ;Jenne, J. ;Kuhlen, V. ;Sinden, D.Günther, M.The emergence of data driven approaches such as Deep Learning has led to novel application of various aspects of science and engineering. It has recently entered the field of ultrasound image beamforming. In this work we investigate neural networks tailored to create images of the quality of multiple compounded plane wave excitations from the data of the central angle (0°) excitation only. The proposed network is used to produce pixel-wise weights to weigh a standard delay-and-sum image from all channel data available to a pixel. It is found to produce higher quality images than the classical reference reconstruction from the 0° angle data.
PublicationDeep learning-based reconstruction of ultrasound images from raw channel data( 2020)
;Strohm, H. ;Rothlübbers, S. ;Eickel, K.Günther, M.Purpose We investigate the feasibility of reconstructing ultrasound images directly from raw channel data using a deep learning network. Starting from the raw data, we present the network the full measurement information, allowing for a more generic reconstruction to form, as compared to common reconstructions constrained by physical models using fixed speed of sound assumptions. Methods We propose a U-Net-like architecture for the given task. Additional layers with strided convolutions downsample the raw data. Hyperparameter optimization was used to find a suitable learning rate. We train and test our deep learning approach on plane wave ultrasound images with a single insonification angle. The dataset includes phantom as well as in vivo data. Results The images produced by our method are visually comparable to ones reconstructed with the conventional delay and sum algorithm. Deviations between prediction and ground truth are likely to be related to speckle noise. For the test set, the mean absolute error is 4.23±1.52 for the phantom images and 6.09±0.72 for the in vivo data. Conclusion The result shows the feasibility of our approach and opens up new research directions regarding information retrieval from raw channel data. As the networks reconstruction performance is limited by the quality of the ground truth images, using other ultrasound reconstruction technique or image types as target data would be of interest.
PublicationLiquid-phase deposition of TiO2 on polystyrene latex particles functionalized by the adsorption of polyelectrolytes( 2005)
;Strohm, H.Löbmann, P.The adsorption of polyelectrolyte layers on polystyrene latex particles was used as a method of surface functionalization for the nucleation and growth of a TiO2 shell on the lattices by liquid-phase deposition. Surface modification by adsorption of a single layer or a bilayer of polyelectrolytes was traced by zeta potential measurements. The TiO2 film formation is very sensitive to the nature of the adsorbed polyelectrolytes, and material deposition in an aqueous latex dispersion ranges from an exclusively homogeneous precipitation of TiO2 particles to a complete coating of the polymer surface with a smooth inorganic shell. An incomplete, patchy TiO2 deposition at the early stage of the coating process leads to a rapid growth of uneven shells, whereas uniform coverage of the substrate at the initial stage leads to the growth of smooth shells at a reduced deposition rate, suggesting that the shell growth rate and the extent of homogeneous precipitation is dependent on the interaction of the substrate with the deposition solution.
PublicationPorous TiO2 hollow spheres by liquid phase deposition on polystyrene latex-stabilised pickering emulsions( 2004)
;Strohm, H.Löbmann, P.
PublicationAssembly of hollow spheres by templated liquid phase deposition following the principles of biomineralisation( 2004)
;Strohm, H.Löbmann, P.Hollow spheres with a hierarchically ordered wall structure are obtained by mineralisation of TiO2 on spherical assemblies of 2 µm polystyrene latex particles in a process mimicking all four stages of natural biomineralisation.
PublicationAdjustment of the band gap energies of biostabilized CdS nanoparticles by application of statistical design of experiments( 2004)
;Barglik Chory, C. ;Remenyi, C. ;Strohm, H.Müller, G.The colloidal synthesis of CdS nanoparticles with the biostabilizers cysteine and glutathione, respectively, at pH values ranging from 4 to 10 is described. For the adjustment of their UV/Vis absorption properties and hence their band gap energies, the Statistical Design of Experiments (DoE) was used. This method allows the simultaneous variation of the synthesis parameters in a systematic manner, and thereby synergistic interaction effects can be obtained. The band gap energies of the quantum dots can be tuned from 3.32 to 4.26 eV by varying kind and concentration of stabilizer, pH value, and concentration of sulfide source. The energy position is significantly dependent on the interaction between the pH value and the concentration of the stabilizer, and the effect of high glutathione concentration is opposite at acidic and alkaline conditions thus leading to band gaps of 4.10 eV at pH = 6 and of 3.64 eV at pH = 10. Examples for the synthesis of semiconductor nanoparticles with predefined spectroscopic properties and preset preparation conditions, e.g., alkaline conditions for the implementation of acid-sensitive dopants, are given.