An AI-based framework for visualizing and analyzing massive amounts of 4D tomography data for beamline end users (KI4D4E)

Third Party Funds Group - Overall project


Acronym: KI4D4E

Start date : 01.03.2023

End date : 28.02.2026

Website: https://foerderportal.bund.de/foekat/jsp/SucheAction.do?actionMode=view&fkz=05D23WE1


Project details

Short description

Synchrotron tomography, using brilliant X-rays, allows for high-resolution, artifact-free imaging. Unlike traditional lab CT, this technique can also conduct rapid 4D tomography experiments, leading to vast amounts of data. The KI4D4E project aims to develop AI methods to process these large data sets. The focus is on artifact reduction, segmentation, and visualization of large 4D data sets. The methods are intended to be applicable to data from both photon and neutron sources.

Scientific Abstract

Synchrotron tomography is characterized by extremely brilliant X-rays, which enables almost artifact-free imaging. Furthermore, very high resolution can be achieved by using special X-ray optics, and the special design of synchrotron facilities also allows fast in-situ experiments, i.e. 4D tomography.  The combination of these features enables high-resolution computed tomography on objects where conventional laboratory CT fails. At the same time, however, this also produces enormous amounts of data that are generally unprocessable by end users, pushing even the operators of synchrotrons to their limits.

The goal of the KI4D4E project is to develop AI-based methods that can be used by end users to process the enormous amounts of data in such 4D CT measurements. This includes improving image quality through artifact reduction, reduction and accessibility of data to end users to help the latter interpret the results.

The project focuses on the topics of artifact reduction, segmentation and visualization of large 4D data sets. The resulting methods should be applicable to data from both photon and neutron sources.

Involved:

Contributing FAU Organisations:

Funding Source

Research Areas