Theoretical variance analysis of single- and dual-energy computed tomography methods for calculating proton stopping power ratios of biological tissues

Abstract

We discovered an empirical relationship between the logarithm of mean excitation energy (ln Im) and the effective atomic number (EAN) of human tissues, which allows for computing patient-specific proton stopping power ratios (SPRs) using dual-energy CT (DECT) imaging. The accuracy of the DECT method was evaluated for 'standard' human tissues as well as their variance. The DECT method was compared to the existing standard clinical practice-a procedure introduced by Schneider et al at the Paul Scherrer Institute (the stoichiometric calibration method). In this simulation study, SPRs were derived from calculated CT numbers of known material compositions, rather than from measurement. For standard human tissues, both methods achieved good accuracy with the root-mean-square (RMS) error well below 1%. For human tissues with small perturbations from standard human tissue compositions, the DECT method was shown to be less sensitive than the stoichiometric calibration method. The RMS error remained below 1% for most cases using the DECT method, which implies that the DECT method might be more suitable for measuring patient-specific tissue compositions to improve the accuracy of treatment planning for charged particle therapy. In this study, the effects of CT imaging artifacts due to the beam hardening effect, scatter, noise, patient movement, etc were not analyzed. The true potential of the DECT method achieved in theoretical conditions may not be fully achievable in clinical settings. Further research and development may be needed to take advantage of the DECT method to characterize individual human tissues.

Determination of mean ionization potential using magnetic resonance imaging for the reduction of proton beam range uncertainties: theory and application

“…The accuracy of DECT parameterizations for SPR determination have varied RMSE from 0.12% -0.28% with maximum errors ranging between 0.39% to 0.98% (Taasti et al, 2016) depending on the type of parameterization used when tested in "reference" tissues. Nearly all of these SECT (stoichiometric) and DECT parameterization methods require parameterization to "reference" human tissues (as those of ICRU Report #44) and suffer from increased errors as elemental compositions for particular tissue types deviate away from them (Taasti et al, 2016;Yang et al, 2010;Yang et al, 2012). In practice, errors from SECT and DECT calculation of Im and SPR can be much larger.…”

“…Five-component models of molecules in the human body have been used previously in the study of the human body (Heymsfield, 2005;Wang et al, 1992) and are assumed, in this study, to be sufficient to accurately calculate the mean ionization potential for naturally occurring biological tissues. From the work of Yang et al (Yang et al, 2010;Yang et al, 2012), they found that the primary contributor to uncertainties in an individual's tissues was hydrogen content in soft tissues and calcium content in mineralized/bony tissues. We investigated the hydrogen dependence of Im for water, lipids, carbohydrates and proteins by determining Im in water, lipids such as triglycerides, glucose and glycogen, and the 20 common (+2 highly uncommon) amino acids that compose proteins.…”

Simplified derivation of stopping power ratio in the human body from dual‐energy CT data

“…So far, several procedures for calculating SPR from DECT data have been proposed. [2][3][4][5][6] Yang et al originally demonstrated that the logarithm of I-value, lnI, can be parameterized as two sectional linear functions of Z eff for the standard human tissues. 2 Han et al proposed a DECT method where I-value was estimated directly from the CT numbers, avoiding the estimation of Z eff .…”

“…[2][3][4][5][6] Yang et al originally demonstrated that the logarithm of I-value, lnI, can be parameterized as two sectional linear functions of Z eff for the standard human tissues. 2 Han et al proposed a DECT method where I-value was estimated directly from the CT numbers, avoiding the estimation of Z eff . 5 Taasti et al proposed an empirical parameterization for SPR based directly on the CT numbers in a DECT data set, whereby the intermediate steps of estimating q e and I-value are not required.…”

The impact of CT scan energy on range calculation in proton therapy planning

“…It is well known that the CT number is dependent on the energy spectrum of the scanner 1 , 2 . Cropp et al (3) demonstrated kVp dependence in the CT number.…”