Validation of Two Mathematical Formalisms for Tissue Characterization in Dual Energy Computed Tomography
A Bourque1*, J Carrier2, H Bouchard2, (1) McGill University, Montreal, Quebec, (2) Centre universitaire de l'Universite de Montreal, Montreal, QuebecSU-E-J-182 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: While dose calculations are typically performed using a simplistic correspondence of HU to electron density (ED), recent developments in DECT for radiotherapy could provide significant improvements in characterizing human tissues for such purpose. We aim to compare and validate two DECT mathematical formalisms and evaluate their accuracy in terms of ED and effective atomic number (Z) for radiotherapy applications.
Methods: Two cylindrical phantoms (Catphan 504 and Gammex 467) containing tissue substitutes are scanned with a Philips Gemini GXL CT at 90, 120 and 140 kV. Two mathematical formalisms are developed and implemented using MATLAB, allowing the extraction of ED and effective Z maps of various materials, given a pair of CT images taken at two distinctive energies. The first formalism is based on a parameterization of XCOM cross sections and uses generic photon spectra provided by the manufacturer. The second formalism is based on a stoichiometric calibration of HU and uses experimental data and the substitutes' composition. A novel definition of effective Z is developed for both formalisms.
Results: With the 90-120 kV energy pair, the extraction of relative ED of the Catphan materials leads to a maximum relative error of 5% for the XCOM-based formalism and 2% for the stoichiometric-based formalism. In the instance of the Gammex materials, higher density materials, as bones, present errors up to 31% and 15% respectively.
Conclusion: While the stoichiometric-based formalism demonstrates a clear advantage over the XCOM-based formalism in the analysis of CT data acquired clinically, both yield reasonable accuracy for low-Z elements materials. Conversely, the results for high-Z materials are negatively affected by discontinuities present in photoelectric effect cross sections. An advanced formalism, which would precisely parameterize this effect, is expected to yield improvements in accuracy and lead the way to a successful implantation of DECT in radiotherapy treatment planning.