Click here to


Are you sure ?

Yes, do it No, cancel

On Deriving X-Ray CT Spectra From Beam Hardening Correction Polynomials

D Han*, S Zhang , P Klahr , B Whiting , J Williamson , J O'Sullivan


(Sunday, 7/29/2018) 2:05 PM - 3:00 PM

Room: Room 205

Purpose: To present an efficient and accurate technique for deriving the x-ray CT spectra from the vendor’s beam-hardening (BH) algorithm based on correction parameters embedded in the raw sinogram header file. Such spectra are necessary to implement model-based statistical image reconstruction algorithm and for monitoring beam-quality stability.

Methods: Our technique requires (a) ability to duplicate the vendor’s BH correction on synthetic raw projection data given BH parameters extracted from the sinogram header; (b) capability of simulating the vendor calibration process used to estimate the BH parameters and understanding the data linearization condition (quality function or QF) to be met; and (c) an independently measured spectrum (Y1) at one time point for identifying small deviations from the ideal QF. The measured reference spectra, denoted as Y1 and Y2, were derived, 7 years apart from each other, by fitting the Birch-Marshall spectrum to narrow beam transmission data through the stacks of Al and Cu filters, measured by CT detector array. Once the QF was derived from the Y1 spectrum and BH parameters contemporaneous with the Y2 spectrum were extracted, the method consists of searching for the kVp and effective filtration (mm Al) to identify the raw BM spectrum such that the synthetic raw calibration phantom projections, corrected by the vendor BH algorithm, matches the QF. The derived Y2 spectrum was compared against its independently measured counterpart.

Results: Compared to the measured Y2 90 kVp spectrum, spectra derived from BH correction differed by 1 keV, 0.3 mm, 0.23 keV and 0.04 for kVp, effective Al filtration, mean energy, and root-mean-squared-error, respectively. For the 140 kVp beam, the corresponding differences are 0 keV, 0.7 mm, 0.56 keV and 0.004, respectively.

Conclusion: Our proposed technique leads to accurate and robust x-ray CT spectra, and requires no additional physical measurements.

Funding Support, Disclosures, and Conflict of Interest: Supported in part by grants R01 CA 149-305 and 212-638. We are indebted to Philips for graciously sharing proprietary information on their raw data formats, BH correction algorithm, and BH calibration process


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email