Extreme-scale precision Imaging in Radio Astronomy (EIRA)
Lead Research Organisation:
Heriot-Watt University
Department Name: Sch of Engineering and Physical Science
Abstract
Aperture synthesis by interferometry in radio astronomy is a powerful technique allowing observation of the sky with antennae arrays at otherwise inaccessible angular resolutions and sensitivities. Image formation is however a complicated problem. Radio-interferometric measurements provide incomplete linear information about the sky, defining an ill-posed inverse imaging problem. Powerful computational imaging algorithms are needed to inject prior information into the data and recover the underlying image.
The transformational science envisaged from radio astronomical observations for the next decades has triggered the development of new gigantic radio telescopes, such as the Square Kilometre Array (SKA), capable of imaging the sky at much higher resolution, with much higher sensitivity than current instruments, over wide fields of view. In this context, wide-band image cubes will exhibit rich structure and reach sizes between 1 Terabyte (TB) and 1 Petabyte (PB), while associated data volumes will reach the Exabyte (EB) scale. Endowing SKA and pathfinders with their expected acute vision requires image formation algorithms capable to transform the data and provide the target imaging precision (i.e. resolution and dynamic range), while simultaneously being robust (i.e. addressing calibration and uncertainty quantification challenges), and scalable to the extreme image sizes and data volumes at stake.
The commonly used imaging algorithm in the field, dubbed CLEAN, owes its success to its simplicity and computational speed. CLEAN however crucially lacks the versatility to handle complex signal models, thereby limiting the achievable resolution and dynamic range of the formed images. The same holds for the existing associated calibration methods that need to correct for instrumental and ionospheric effects affecting the data. Another major limitation in radio-interferometric imaging is the absence of a proper methodology to quantify the uncertainty around the image estimate.
A decade of research pioneered by Wiaux and his collaborators suggests that the theory of optimisation is a powerful and versatile framework to design new radio-interferometric imaging algorithms. In the optimisation framework, an objective function is defined as sum of a data-fidelity term and a regularisation term promoting a given prior signal model. Our research hypothesis is that algorithmic structures currently emerging at the interface of optimisation and deep learning can take the challenge of delivering the expected generation of algorithms for precision robust scalable radio-interferometric imaging, in a wide-band wide-field polarisation context.
A novel approach will be developed in this context, based on the decomposition of the data into blocks and of the image cube into small, regular, overlapping 3D facets. Facet-specific regularisation terms and block-specific data-fidelity terms will all be handled in parallel through so-called proximal splitting optimisation methods, thereby unlocking simultaneously the image and data size bottlenecks. Injecting prior information into the inverse imaging problem at facet level also offers potential to better promote local spatio-spectral correlation, and eventually provide the target image precision. Sophisticated prior models based on advanced regularisation simultaneously promoting sparsity, correlation, positivity etc., will firstly be considered, to be substituted by learned priors using deep neural networks in a second stage with the aim to further improve precision and scalability. Facets and neural networks will percolate from the imaging module to calibration and uncertainty quantification for robustness. Our algorithms will be validated up to 10TB image size on High Performance Computing (HPC) machines. A technology transfer at 1GB image size will be performed in medical imaging, specifically 3D magnetic resonance and ultrasound imaging, as proof of their wider applicability.
The transformational science envisaged from radio astronomical observations for the next decades has triggered the development of new gigantic radio telescopes, such as the Square Kilometre Array (SKA), capable of imaging the sky at much higher resolution, with much higher sensitivity than current instruments, over wide fields of view. In this context, wide-band image cubes will exhibit rich structure and reach sizes between 1 Terabyte (TB) and 1 Petabyte (PB), while associated data volumes will reach the Exabyte (EB) scale. Endowing SKA and pathfinders with their expected acute vision requires image formation algorithms capable to transform the data and provide the target imaging precision (i.e. resolution and dynamic range), while simultaneously being robust (i.e. addressing calibration and uncertainty quantification challenges), and scalable to the extreme image sizes and data volumes at stake.
The commonly used imaging algorithm in the field, dubbed CLEAN, owes its success to its simplicity and computational speed. CLEAN however crucially lacks the versatility to handle complex signal models, thereby limiting the achievable resolution and dynamic range of the formed images. The same holds for the existing associated calibration methods that need to correct for instrumental and ionospheric effects affecting the data. Another major limitation in radio-interferometric imaging is the absence of a proper methodology to quantify the uncertainty around the image estimate.
A decade of research pioneered by Wiaux and his collaborators suggests that the theory of optimisation is a powerful and versatile framework to design new radio-interferometric imaging algorithms. In the optimisation framework, an objective function is defined as sum of a data-fidelity term and a regularisation term promoting a given prior signal model. Our research hypothesis is that algorithmic structures currently emerging at the interface of optimisation and deep learning can take the challenge of delivering the expected generation of algorithms for precision robust scalable radio-interferometric imaging, in a wide-band wide-field polarisation context.
A novel approach will be developed in this context, based on the decomposition of the data into blocks and of the image cube into small, regular, overlapping 3D facets. Facet-specific regularisation terms and block-specific data-fidelity terms will all be handled in parallel through so-called proximal splitting optimisation methods, thereby unlocking simultaneously the image and data size bottlenecks. Injecting prior information into the inverse imaging problem at facet level also offers potential to better promote local spatio-spectral correlation, and eventually provide the target image precision. Sophisticated prior models based on advanced regularisation simultaneously promoting sparsity, correlation, positivity etc., will firstly be considered, to be substituted by learned priors using deep neural networks in a second stage with the aim to further improve precision and scalability. Facets and neural networks will percolate from the imaging module to calibration and uncertainty quantification for robustness. Our algorithms will be validated up to 10TB image size on High Performance Computing (HPC) machines. A technology transfer at 1GB image size will be performed in medical imaging, specifically 3D magnetic resonance and ultrasound imaging, as proof of their wider applicability.
Planned Impact
Firstly, EIRA will bring new advances for precision robust scalable computational imaging, with primary application to radio astronomy. By bringing new knowledge from a multi-disciplinary perspective (ranging from theoretical aspects in optimisation and deep learning, to astronomical imaging, and high performance computing), our results will induce paradigm shifts that would be unreachable by the radio astronomy, signal processing, and high performance computing (HPC) communities separately. This direct academic impact will contribute to give SKA and precursors the acute vision they require, thereby indirectly contributing in the longer term to their expected scientific impact (advancing astrophysics), economic impact (technological developments to build the instrument and to process unprecedented amounts of data) and societal impact (understanding the origin of life, etc.).
Secondly, while the project will focus on radio astronomy, research time is allocated to transfer results to medical imaging, in particular to magnetic resonance (MR) and ultrasound (US) imaging, ensuring additional direct academic impact. Our approach will open the door to more robust diagnostic methodologies at early stages of diseases and therefore at lower cost, which constitutes an indirect economic and societal impact on healthcare in the longer term.
The collaboration with Siemens Healthineers, who will provide regular feedback and evaluation of our developments for MR imaging, is an essential pathway towards realisation of this impact. This collaboration will provide Siemens with a direct economic advantage in having prime access to our developments and software platform. Further industrial relations will be sought with the aim to further open the areas of impact of our developments, on the medical imaging market beyond MR and US imaging, and on the defence market (e.g. Canon Medical Research, Leonardo, Seebyte). Our novel imaging technologies will be patented if suitable, which will represent a first step in a commercialisation perspective.
Thirdly, the planned public conferences will contribute to dissemination of knowledge, which represents a first stage of societal impact on the project timescale.
Finally, EIRA will have a direct impact on people on the project timescale. The research associates will sharpen their skills for the mathematical formulation of complex problems and for analytic reasoning. Beyond these hard (i.e. technical) skills, they will also acquire essential soft (i.e. interpersonal) skills for management, leadership, team work and communication. This portfolio of skills will add to their professional development and help them become leaders in any employment sector.
Secondly, while the project will focus on radio astronomy, research time is allocated to transfer results to medical imaging, in particular to magnetic resonance (MR) and ultrasound (US) imaging, ensuring additional direct academic impact. Our approach will open the door to more robust diagnostic methodologies at early stages of diseases and therefore at lower cost, which constitutes an indirect economic and societal impact on healthcare in the longer term.
The collaboration with Siemens Healthineers, who will provide regular feedback and evaluation of our developments for MR imaging, is an essential pathway towards realisation of this impact. This collaboration will provide Siemens with a direct economic advantage in having prime access to our developments and software platform. Further industrial relations will be sought with the aim to further open the areas of impact of our developments, on the medical imaging market beyond MR and US imaging, and on the defence market (e.g. Canon Medical Research, Leonardo, Seebyte). Our novel imaging technologies will be patented if suitable, which will represent a first step in a commercialisation perspective.
Thirdly, the planned public conferences will contribute to dissemination of knowledge, which represents a first stage of societal impact on the project timescale.
Finally, EIRA will have a direct impact on people on the project timescale. The research associates will sharpen their skills for the mathematical formulation of complex problems and for analytic reasoning. Beyond these hard (i.e. technical) skills, they will also acquire essential soft (i.e. interpersonal) skills for management, leadership, team work and communication. This portfolio of skills will add to their professional development and help them become leaders in any employment sector.
Organisations
- Heriot-Watt University (Lead Research Organisation)
- Science and Technology Facilities Council (Co-funder)
- Siemens (United Kingdom) (Project Partner)
- CentraleSupélec (Project Partner)
- National Radio Astronomy Observatory (Project Partner)
- École Polytechnique Fédérale de Lausanne (Project Partner)
- Rhodes University (Project Partner)
- Square Kilometre Array Organisation (Project Partner)
Publications
Aghabiglou A
(2023)
Deep Network Series for Large-Scale High-Dynamic Range Imaging
Dabbech A
(2022)
First AI for Deep Super-resolution Wide-field Imaging in Radio Astronomy: Unveiling Structure in ESO 137-006
in The Astrophysical Journal Letters
Dabbech A
(2021)
Cygnus A jointly calibrated and imaged via non-convex optimization from VLA data
in Monthly Notices of the Royal Astronomical Society
Eldaly AK
(2021)
Bayesian Activity Estimation and Uncertainty Quantification of Spent Nuclear Fuel Using Passive Gamma Emission Tomography.
in Journal of imaging
Pesce M
(2021)
Fast Fiber Orientation Estimation in Diffusion MRI from kq-Space Sampling and Anatomical Priors.
in Journal of imaging
Pesquet J
(2020)
Learning Maximally Monotone Operators for Image Recovery
Pesquet J
(2021)
Learning Maximally Monotone Operators for Image Recovery
in SIAM Journal on Imaging Sciences
Repetti A
(2021)
Variable Metric Forward-Backward Algorithm for Composite Minimization Problems
in SIAM Journal on Optimization
Repetti A
(2022)
Dual Forward-Backward Unfolded Network for Flexible Plug-and-Play
Tasse C
(2021)
The LOFAR Two-meter Sky Survey: Deep Fields Data Release 1 I. Direction-dependent calibration and imaging
in Astronomy & Astrophysics
Description | The combination of new Artificial Intelligence and High Performance Computing technologies enable both improving the precision of image reconstruction algorithms for high resolution high dynamic range imaging, and ensuring reasonable compution times. |
Exploitation Route | The project will also investigate a transfer of technology towards medical imaging |
Sectors | Digital/Communication/Information Technologies (including Software) Healthcare |
Title | BASPLib |
Description | BASPLib is an open-source library on GitHub, gathering Python and Matlab codes to solve challenging inverse imaging problems in astronomy and medicine. The primary imaging modality of focus is synthesis imaging by interferometry in radio astronomy, with functionality currently being developed for magnetic resonance imaging and ultrasound imaging in medicine. The BASPLib software suite gathers implementations of the most advanced computational imaging algorithms at the interface of optimisation and deep learning theories. The proposed algorithms can be seen as intermediate steps in the quest for an ultimate "intelligent" imaging algorithm (yet to be devised) providing the joint precision, robustness, efficiency, and scalability required by modern applications. A key feature on this past is algorithm modularity, with regularisation modules (enforcing image and calibration models) alternating with data-fidelity modules (enforcing consistency with the observed data). BASPLib algorithms and software are developed at Edinburgh's Biomedical and Astronomical Signal Processing Laboratory (https://basp.site.hw.ac.uk/) headed by Prof. Wiaux. |
Type Of Technology | Software |
Year Produced | 2024 |
Open Source License? | Yes |
Impact | Makes advance computational imaging algorithms available to the community and recently triggered new algorithmic developments in astronomical imaging. |
URL | https://basp-group.github.io/BASPLib/ |