Improving reproducibility in synchrotron tomography using implementation-adapted filters

TL;DR

This paper presents a method to improve the reproducibility of synchrotron tomography reconstructions by optimizing filters in analytical algorithms, ensuring consistent results across different software implementations and datasets.

Contribution

It introduces a technique to compute implementation-adapted filters that reduce differences caused by discretisation and interpolation in reconstruction algorithms.

Findings

Implementation-adapted filters lead to more consistent reconstructions.

The method works on both simulated and real-world data.

It enhances reproducibility across different software packages.

Abstract

For reconstructing large tomographic datasets fast, filtered backprojection-type or Fourier-based algorithms are still the method of choice, as they have been for decades. These robust and computationally efficient algorithms have been integrated in a broad range of software packages. The continuous mathematical formulas used for image reconstruction in such algorithms are unambiguous. However, variations in discretisation and interpolation result in quantitative differences between reconstructed images, and corresponding segmentations, obtained from different software. This hinders reproducibility of experimental results, making it difficult to ensure that results and conclusions from experiments can be reproduced at different facilities or using different software. In this paper, we propose a way to reduce such differences by optimising the filter used in analytical algorithms. These filters can be computed using a wrapper routine around a black-box implementation of a reconstruction algorithm, and lead to quantitatively similar reconstructions. We demonstrate use cases for our approach by computing implementation-adapted filters for several open-source implementations and applying it to simulated phantoms and real-world data acquired at the synchrotron. Our contribution to a reproducible reconstruction step forms a building block towards a fully reproducible synchrotron tomography data processing pipeline.