CT reconstruction using event-driven data from Timepix4

The project will cover mathematical, computational and experimental aspects. Reference data is available, but new measurements during the project are expected.

The project is a collaboration between the Computational Imaging Group at CWI (T.van.Leeuwen@cwi.nl) and the Amsterdam Scientific Instruments (ASI) (erik.hogenbirk@amscins.com).

Project description

ASI develops and produces X-ray detectors based on CERN-developed Timepix4 chips. These detectors can be used for CT reconstruction. Sample measurements have recently been performed, which demonstrate the performance of such a system for applications such as battery inspection or PCB solder quality inspection.

The nature of data from the detector is quite unique. Ordinary X-ray detectors produce an image for a fixed exposure time, so that an image is produced with a value for the intensity for each pixel. In contrast, the pixels in Timepix4 each individually record an event when an X-ray photon is measured. For each of those interactions, the position, time (sub-nanosecond precision) and energy are measured. Images can be made by producing 2D histograms of the (x,y) position of the events, but more complex analyses are possible with this data.

The nature of the project is twofold:

  1. Investigate the optimal way of using the event-driven data for CT reconstruction. Possibilities include X-ray energy selection for the image creation, sub-pixel resolution imaging, or drift correction for continuous rotation experiments. Alternatively, an event-based (sparse) algorithm could be developed that skips the 2D image step entirely, thereby optimizing the approach for event-based data.
  2. New measurements with a Timepix4 camera from ASI in the CT setup at CWI are foreseen. Samples could include cultural heritage objects. These measurements could demonstrate the performance of the reconstruction methods developed or allow new insights based on the samples available.

FlexRay-scanner at CWI scanning an object and two images of CT-scans of the object..

Supervision & focus areas

Supervisors: Tristan van Leeuwen, Erik Hogenbirk
Keywords: tomographic imaging, sparse data, Timepix4