Penumbra-effect induced spectral mixing in x-ray computed tomography: a multi-ray spectrum estimation model and...

Area of Science:

• Medical Physics
• Biomedical Imaging
• Computational Imaging

Background:

• Spectral CT imaging utilizes various filter-based methods for enhanced diagnostic capabilities.
• Spectral filters can introduce spectral mixing in the penumbra region due to focal spot size, reducing accuracy.
• Traditional spectrum estimation methods do not account for this penumbra effect.

Purpose of the Study:

• To develop a multi-ray spectrum estimation model to address spectral mixing in the penumbra region.
• To propose the Adaptive Subsampled WeIghting of Filter Thickness (A-SWIFT) method for improved spectral accuracy.
• To enhance the performance of CT systems employing spectral filters.

Main Methods:

• Estimated unfiltered spectra using traditional methods.
• Modeled final spectra as a weighted summation of multi-filter attenuated spectra.
• Obtained weights and equivalent lengths via x-ray transmission measurements with altered spectra (kVp or flat filters).
• Approximated spectra using the multi-ray model and mimicked penumbra effects with spectral modulators and split filters in simulations and experiments.

Main Results:

• Reduced mean energy bias in the penumbra region from 7.43 keV to 0.72 keV (split filter) and 1.98 keV to 0.61 keV (spectral modulator) compared to previous methods.
• Decreased average error of mean values (ERMSE) from 77 to 7 Hounsfield units (HU) for a pure water phantom.
• Improved iodine concentration estimation accuracy from 1.5 mg/ml to 0.2 mg/ml for a Gammex phantom.

Conclusions:

• The A-SWIFT method, based on a multi-ray spectrum estimation model, offers accurate and robust spectral estimation in the penumbra region.
• This method significantly enhances spectral imaging performance in CT systems using spectral filters.
• A-SWIFT contributes to more precise material decomposition and quantitative imaging in spectral CT.