Quantitative analysis of metal artifacts in X-ray tomography

TL;DR

This paper provides a detailed microlocal analysis of streaking artifacts in X-ray CT scans caused by metallic objects, identifying their mathematical nature and exploring artifact reduction techniques.

Contribution

It offers a novel quantitative microlocal analysis of metal artifacts in X-ray tomography, including effects of boundary geometry and artifact mitigation methods.

Findings

Artifacts are conormal distributions to lines tangent to boundary curves.

Corner points significantly influence streaking artifacts.

Filtering can reduce the severity of artifacts.

Abstract

In X-ray CT scan with metallic objects, it is known that direct application of the filtered back-projection (FBP) formula leads to streaking artifacts in the reconstruction. These are characterized mathematically in terms of wave front sets in [13]. In this work, we give a quantitative microlocal analysis of such artifacts. We consider metal regions with strictly convex smooth boundaries and show that the streaking artifacts are conormal distributions to straight lines tangential to at least two boundary curves. For metal regions with piecewise smooth boundaries, we analyze the streaking artifacts especially due to the corner points. Finally, we study the reduction of the artifacts using appropriate filters.