Encrypted login | home

Program Information

Beam Hardening Correction Using Eigentissue Decomposition On DECT Raw Data

no image available
M Simard

M Simard1*, H Bouchard1,2 , (1) Universite de Montreal, Montreal, Canada, (2) CRCHUM, Montreal, Canada.


WE-DE-605-6 (Wednesday, August 2, 2017) 10:15 AM - 12:15 PM Room: 605

Purpose: Several dual-energy CT (DECT) tissue characterization methods were recently proposed. Because these methods are post-reconstruction, they are limited by the accuracy of semi-empirical beam hardening correction methods. The purpose of this work is to demonstrate the potential of a novel pre-reconstruction approach to correct for beam hardening effects and provide accurate tissue characterization for radiotherapy.

Methods: A new sinogram-based DECT formalism based on eigentissue decomposition (ETD) (Lalonde and Bouchard, 2016) is proposed. The method allows reconstructing elemental fraction and electron density (ED) maps. Raw-data ETD is compared with two sinogram-based DECT methods: Alvarez and Macovski (AM) and two-material decomposition (2MD). Elemental fractions are evaluated using the parametrization of H√ľnemohr (2014). A DECT simulation framework is developed to compare tissue parameters, stopping power relative to water (SPR) and photon energy attenuation coefficient (EAC) across formalisms for various noise and spectrum filtration levels.

Results: Validation against other formalisms show that raw-data ETD efficiently removes beam hardening effects. Considering all noise levels (0-15 HU), ETD outperforms the other formalisms, providing lower RMS errors for elemental fractions. For SPR, RMS errors for ETD ranges from 0.13-0.92%, as opposed to 0.57-0.89% (2MD) and 0.73-0.97% (AM), while the bias on SPR respectively ranges from 0.03-0.04%, 0.45-0.50% and 0.65-0.67%. For EAC, RMS errors are between 0.7-12.2%, 6.1-11.8% and 11.6-15.3%. The bias on EAC ranges from 0.1-0.2%, 4.4-4.9% and 11-11.5%. For all spectrum alteration levels, ETD is more accurate and precise than 2MD and AM.

Conclusion: This work shows the potential of raw-data based ETD in providing precise, accurate and beam hardening-free tissue characterization parameters. For all noise and spectrum alteration levels considered, ETD is more robust and provides more accurate tissue chemical compositions and quantities related to dose calculation (SPR, EAC). ETD is also generalizable to more than two energies and possibly applicable to spectral CT.

Funding Support, Disclosures, and Conflict of Interest: MS acknowledges fellowship support from the Natural Sciences and Engineering Research Council (NSERC).

Contact Email: