University of Twente
Location: The Netherlands
3 PhD positions on super-resolution ultrasound imaging in the university of Twente (The Netherlands).
(ERC starting grant 2022 number 101076844 )
Research Project and timeline
Super-FALCON is a 1.88 M€ project funded by the European Research Council (funding instrument: ERC Starting Grant) on Super-resolution, ultrafast and deeply-learned contrast ultrasound imaging of the vascular tree. Starting dates for the proposed positions are between May 2023 and October 2023 as per agreement between the candidate and the local supervisor. The positions will be closed as soon as a suitable candidate is found (which can be before theagreed-upon starting date of the contract). The duration of the project is 4 years.
Context and general project description
Our healthcare system is under unsustainable strain owing, largely, to cardiovascular diseases and cancer. For both, imaging vasculature and flow precisely is paramount to reduce costs while improving diagnosis and treatment. Specifically, the focus is on the multiscale aspects of shear, vorticity, pressure, and mechanics within arteries and capillary beds (10 – 200 μm vessels). However, this requires an imaging depth of ~10 cm with a resolution of ~50 μm. Furthermore, velocities often exceed 1m/s, which requires a frame rate of ~1000 fps. So far, clinical imaging modalities have been incapable of reaching sufficient spatiotemporal resolution and there is thus a dire need for new techniques.
Plane-wave ultrasound enhanced with contrast microbubbles outperforms all modalities in safety, cost, and speed, and is thus the ideal candidate to address this need. The strategy proposed in the Super-FALCON project harnesses the nonlinear dynamics of monodisperse microbubbles. Work package 1 (PhD position 1) will use deep learning and GPU-accelerated acoustic simulations to recover super-resolved (1/20th of the wavelength) bubble clouds. In work package 2 (PhD position 2), we will create a new physical model for confined bubbles and use them as nonlinear sensors for capillary imaging. In work package 3 (PhD position 3), we will disentangle attenuation and scattering using the physics and mathematics of nonlinear ultrasound propagation complemented with deep learning. This will allow us to correct for wave distortion and apply the strategies from work packages 1 and 2 in deep tissue. Finally, in work package 4 (postdoc position), we will use automatic segmentation to integrate the results of WP1, 2 and 3 into a technology that we will scientifically assess on vascularized ex vivo livers.
The ambition of the Super-FALCON project team is to generate a long-term impact both scientifically and societally though new fundamental knowledge (confined bubble dynamics, inhomogeneous ultrasound propagation, and deconvolution strategies) and through new experimental methods for flow imaging and characterization. Carried by the right team, Super-FALCON could initiate a paradigm shift towards patient-specific treatments.
Short description – Work package 1 (PhD position 1)
In this project, you will build a new arterial flow imaging strategy using nonlinear bubble dynamics. You will bring an innovative neural network  to the next level in order to identify and isolate bubble signals within the raw ultrasound data received by the ultrasound transducer. These bubble signatures are nonlinear and, in mathematical terms, form an inhomogeneous convolution kernel. You will also directly integrate image reconstruction in the network to provide accurate representations of dense bubble clouds at a resolution of 1/20th of the US wavelength. You will exploit new optimal transport algorithms to translate super-resolved images into real-time velocity and pressure gradient fields. To generate the training data, you will be able to use a novel simulator (developed in collaboration with the university of Delft) which couples flow dynamics, inhomogeneous propagation, and nonlinear bubble physics to generate the training data. You will first assess your discoveries on a rotating drum flow setup, which provides known velocity and pressure fields applied to bubbles in free field. Validation in pulsatile flow will be conducted in our in-vivo-mimicking setup.
 Blanken, N. , Wolterink, J. M., Delingette, H. , Brune, C. , Versluis, M. , & Lajoinie, G. (2022). Super-Resolved Microbubble Localization in Single-Channel Ultrasound RF Signals Using Deep Learning. IEEE transactions on medical imaging, 41(9), 2532-2542. https://doi.org/10.1109/TMI.2022.3166443
Short description – Work package 2 (PhD position 2)
Work package 2 consists in a balanced amount of physical modeling and experimental work. Your first task will be to develop a model for bubbles confined in viscoelastic vessels. This model must be both simple enough to be easy to handle and accurate enough to represent the reality we can measure. To this end, you will benefit from the extensive experience of the group in modeling bubble dynamics in complex cases [1-3]. Your second task (effectively taking place concurrently) will be to exploit the effect of confinement on bubble echoes to experimentally measure vessel size and elasticity. You will be able to use the neural network developed by your colleague in work package 1 to locate and isolate the individual bubble signals. Together, we will go further and design nonlinear pulsing scheme to readily measure single microbubbles in their minute details. Experimental validation will be conducted on perfused capillary phantoms (10 to 200 μm diameter) fabricated by molding micro-wires in tissue-mimicking hydrogels.
 Versluis, M. et al. Ultrasound Contrast Agent Modeling: A Review. Ultrasound Med. Biol. 46, 2117-2144 (2020).
 Lajoinie, G. et al. Non-spherical oscillations drive the ultrasound-mediated release from targeted microbubbles. Commun. Phys. 1, 22 (2018).
 Wang, Y. et al. Giant and explosive plasmonic bubbles by delayed nucleation. PNAS 115, 7676 (2018).
Short description – Work package 3 (PhD position 3)
Work packages 1 and 2 require accurate sensing of the bubble responses. Distortion caused by frequency-dependent attenuation of the ultrasound waves must therefore be corrected. The main difficulty lies in the entanglement of attenuation and local random scattering that are convolved in the raw ultrasound data. Your work in this work package will therefore be instrumental to the success of the project: you will use the newest developments in deep learning (e.g., physics-informed AI and multiplicative filter networks) to separate the contributions of attenuation and scattering based on the distinct effect they have on the frequency content of the signal. Once established, these developments will allow for recovering the signals actually emitted by the microbubbles and effectively enable the practical application of the results from works packages 1 and 2.
You will be embedded in the Physics of Fluids group (https://pof.tnw.utwente.nl/) of the University of Twente (https://www.utwente.nl/en/), and part of the TechMed Center(https://www.utwente.nl/en/techmed/). This will give you access to state-of-the-art facilities and scientific knowledge. Beyond your direct collaborators of the project, you will also interact with direct colleagues in the group on a scientific level, with the technical staff that will support you in designing efficient experimental setups (for example), and with the applied mathematics group for additional support in computer science and deep learning. You will therefore have all the guidance you may require in your local environment, as well as the freedom to apply your own ideas and intuitions.
– (For WP 1 and 3) You have at some experience with deep learning e.g., during your training and/or internships
Information and application
Your reaction should include an application/motivation letter, emphasizing your specific interest and motivation, a detailed CV, and an academic transcript of B.Sc. and M.Sc. education. Please apply by email to firstname.lastname@example.org. An interview and a scientific presentation will be part of the selection procedure. For more information about the position, you are encouraged to contact Guillaume Lajoinie ( email@example.com )