Regression fitting megavoltage depth dose curves to determine material relative electron density in radiotherapy.

Abstract

Abstract The relative electron density (RED) parameter is ubiquitous throughout radiotherapy for clinical dosimetry and treatment planning purposes as it provides a more accurate description of the relevant radiological properties over mass density alone [1]. RED is theoretically derived and in practice determined indirectly from calibrated CT Hounsfield Units (HU) [1]. While CT images provide useful 3D information, the spectral differences between CT and clinical LINAC beams may impact the validity of the CT-ED calibration, especially in the context of novel tissue-mimicking materials where deviations from biologically typical atomic number to atomic weight ratios < Z/A > occur and/or high-Z materials are present [2]. A theoretical basis for determining material properties directly in a clinical beam spectrum via an electron-density equivalent pathlength (eEPL) method has been previously established [2]. An experimental implementation of this approach is introduced whereby material-specific measured percentage depth dose curves (PDDs) are regressed to a PDD measured in a reference material (water), providing an inference of < Z/A>, which when combined with the physical density provides a determination of RED. This method was validated over a range of tissue-mimicking materials and compared against the standard CT output, as well as compositional information obtained from the manufacturer's specifications. The measured PDD regression method showed consistent results against both manufacturer-provided values and CT-derived within 0.9 and 1.15 RED. Outside of this soft-tissue range a trend was observed whereby the < Z/A > determined becomes unrealistic indicating the method is no longer reporting RED alone and the assumptions around the eEPL model are constrained. Within a range of validity, the regression method provides a practical and robust characterisation for unknown materials in the clinical setting and may be used to improve on the CT output where high Z material components are suspected.