Cone-beam computed tomography reconstruction for a commercial proton beam therapy system

TL;DR

This paper presents a method to reconstruct CBCT images for a proton therapy system, improving image quality by addressing exposure and beam hardening issues.

Contribution

A novel offline CBCT reconstruction workflow for a commercial proton therapy system that accounts for exposure and beam hardening effects.

Findings

The workflow eliminated crescent artifacts in phantom and patient images.

Signal-to-noise and contrast-to-noise ratios were comparable to clinical reconstructions.

The method provides a tool for optimizing acquisition parameters and developing new reconstruction techniques.

Abstract

Cone-beam computed tomography (CBCT) images are used in image-guided radiotherapy to track anatomical changes throughout treatment and to set up patients to ensure accurate delivery of therapeutic radiation at each treatment session. An offline method of CBCT reconstruction workflow, operating on 2D projection images and specific to the imaging system in question, is needed for many image optimisation studies. Here we present a methodology to reconstruct CBCT images from these data for a commercial proton beam therapy machine, accounting for the variation in exposure and beam hardening from filtration due to gantry rotation during CBCT acquisition.

Projection data of solid water phantoms were acquired to model bow-tie filter motion and beam hardening effects. Projection data and system CBCT reconstructions of a Catphan504 phantom were acquired for validation of the method, as well as a retrospectively accessed patient image. The presented workflow was assessed against the clinical reconstructions using uniformity, signal-to-noise-ratio, and contrast-to-noise-ratio measured in the phantom images.

The offline workflow eliminated crescent artefacts due to variable exposure and beam hardening in phantom and patient images. Signal-to-noise and contrast-to-noise ratios were similar compared to system reconstructions, although with slight differences thought to be due to interplay effects in the bow-tie filter.

A workflow was developed to emulate the CBCT reconstruction process for a commercial proton therapy machine, providing a useful tool for optimised acquisition parameters and novel reconstruction processes using this system.

Figures