Advanced tomographic methods for the ESRF EBS upgrade

Start Date
28-11-2018 10:00
End Date
28-11-2018 11:00
Room 500 - 501, Central Building
Speaker's name
Speaker's institute
Computational Imaging Group - Amsterdam,The Netherlands
Contact name
Anaïs Fernandes
Host name
Paul Tafforeau
Add event to calendar
iCal | vCal

The scheduled ESRF EBS upgrade will offer unprecedented opportunities for tomographic applications. In particular, the higher available X-ray photon flux will allow to overcome current limitations, and will introduce new challenges in fields like: phase contrast tomography (PCT), microstructure characterization techniques based on X-ray diffraction (XRD) and X-ray fluorescence (XRF) tomography, as well as dynamic tomography (Dyn-CT). During my seminar, I will discuss how I intend to overcome those limitations and challenges, by bringing to the ESRF research on the latest developments in the field of tomography. These developments include: multi-channel / multi-modal regularization, patterned imaging techniques, and deep-learning.

The increased X-ray flux and the consequent reduction of acquisition times allow to increase the acquisition rate and data throughput of absorption tomography and PCT experiments. The resulting larger datasets will require longer time to be reconstructed, and much larger storage to be stored long term. To cope with this problem, I intend to bring to the ESRF the latest advancements in real-time tomography and data-reduction techniques. These methods both use cutting edge developments in high performance computing (which includes GPU acceleration), but also leverage the rapidly growing field of machine-learning. They will increment data reconstruction / processing speeds, lower the burden on the ESRF networking / computing infrastructure, and greatly reduce data storage requirements.

Reduced acquisition times and real-time tomography will directly benefit Dyn-CT applications, allowing to capture the evolution of rapidly changing features in the analyzed samples.

Induced emission tomography (IET) acquisitions (including XRD-CT and XRF-CT) are normally performed using pencil beam scanning, to render the tomographic problem tractable. This results in long acquisition times. These techniques will greatly benefit from the increased X-ray flux of the EBS upgrade, and will see a much larger interest and applicability.

Both XRD and XRF produce multi-channel data, but are commonly analyzed in a single-channel manner. Moreover, they can often be combined in the same acquisition, providing complementary information about the sample microstructure. In this context, I intend to bring the following research topics to the ESRF: multi-channel regularization and reconstruction; multi-modal reconstructions; and patterned illumination reconstructions.

Multi-channel and multi-modal regularization will allow to both extract higher quality reconstructions from multi-channel acquisitions, and to better correlate the volumes obtained through different techniques (modalities).

The broad applicability of these methods will also benefit a large spectrum of other applications, including for instance: deformation reconstruction using XRD-CT pencil beam acquisitions; and  orientation distribution function (ODF) reconstruction from powder-like data.Finally, software support for patterned illumination reconstructions will allow to leverage the latest advancements in compressive / ghost imaging, reducing the deposited dose, and further reducing the acquisition times.

Visitors from off-site please contact Anaïs Fernandes tel +33 (0)438 88 19 92 to arrange for a gate pass.
Requests made by e-mail will be confirmed.
If you do not receive a confirmation e-mail, please contact us by phone.