The application of dual-energy CT within the U.K. proton beam therapy service

Abstract

The accurate delivery of proton beam therapy (PBT) relies upon the prediction of stopping power ratio (SPR). Treatment planning for PBT requires a CT scan of the patient from which the Hounsfield units (HU) in the scan are related to SPR for the purposes of dose calculation. The current solution for this calibration uses a single-energy CT (SECT) scan and a biologically optimised calibration process accounting for the influence of elemental composition on the SPR of different tissues. Dual-energy CT (DECT) has been proposed as an alternative method of CT calibration which offers the potential for improved SPR prediction accuracy. In this work an empirical method of DECT calibration first developed by Taasti et al. (2016) was adapted for use in the context of the PBT Centre at The Christie NHS Foundation Trust using a Siemens SOMATOM Confidence CT scanner (Siemens Healthineers, Erlangen, Germany). The empirical DECT calibration method was implemented using in-house developed Python code to produce an SPR map from two sequentially acquired CT scans at 80 and 140 kVp. It was validated theoretically using two sets of reference human tissue compositions compared to a ground truth SPR calculation. Experimental validation was also performed using animal tissue samples. Residual range measurements using a multi-layer ionisation chamber were compared with predicted ranges using the existing SECT calibration technique and the proposed DECT methodology. A thorough sensitivity analysis of the DECT calibration was also performed whereby several variables were investigated to determine their effect on the predicted SPR in the animal tissue samples. In ICRP reference human tissues (ICRP, 2009), the DECT calibration predicted SPR with a root-mean-square (RMS) error of 0.51% compared to 0.62% with the SECT solution. In the second reference set of tissues (White et al., 1987) the RMS error with DECT was 0.66% compared to 1.37% with SECT. Experimental validation yielded a mean error in residual proton range of 0.77 mm and an RMS error of 2.26 mm with DECT compared to -0.95 mm (mean) and 2.91 mm (RMS) with SECT. Whilst the overall results indicate that the DECT calibration gives comparable results to SECT, DECT was not consistently superior to SECT in predicting SPR. Beam hardening in the CT scan both at the calibration and imaging stage has an impact on the accuracy of the predicted SPR. This sensitivity can be managed by choosing an appropriate calibration phantom size and by using a beam hardening correction when reconstructing the DECT images. With the Siemens SOMATOM Confidence scanner, scans at different energies are acquired sequentially. The DECT calibration was found to be most sensitive to movement of the patient (simulated with a computational phantom) between these scans which introduced significant errors into the prediction of SPR. This acquisition limitation should be addressed with appropriate immobilisation and careful patient selection prior to clinical implementation. This work has demonstrated that DECT is feasible and practical in the U.K. PBT service and an empirical methodology has been adapted to suit the commercially available equipment that is in use. The method has been validated by measurement in the form of a complete end-to-end test using biological samples. The limitations to the methodology have been investigated by a thorough sensitivity analysis to ensure that clinical risks can be mitigated prior to any future implementation.

Date of Award	31 Dec 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Marianne Aznar (Supervisor)