Author: Giavanna Luisa Jadick, Patrick J La Riviere 👨🔬
Affiliation: University of Chicago 🌍
Purpose: We assess two multi-measurement acquisition schemes for material decomposition with x-ray phase-contrast imaging (XPCI); demonstrating for the first time that multi-distance imaging can match or outperform traditional multi-energy imaging.
Methods: In propagation-based XPCI material decomposition, a system of equations is solved for each frequency component of the input images. Without spectral information, the equations are algorithmically unstable at the 0-frequency; thus, traditionally they use multi-energy inputs. Here we also examine performance using multi-distance inputs, while being mindful of the potential for a singular system of equations. We simulated XPCI acquisitions with a range of energies (15–35 keV), propagation distances (0–200 mm), and doses (103–105 photons/pixel). A photon-counting detector was modeled with 8-micron pixels, Lorentzian point-spread function, and Poisson noise. The Shepp–Logan phantom was imaged, modified to comprise only bone and soft tissue. Multi-distance and multi-energy material decomposition were performed using all combinations of images acquired with two distances or two energies, respectively.
Results: Average structural similarity index (SSIM) and root-mean-square-error (RMSE) were measured in each pair of material images. At low dose, multi-distance achieved a better maximum SSIM (0.70 versus 0.46) and minimum RMSE (0.09 versus 0.19). At higher dose, the two techniques achieved similar optimized SSIM (0.81 versus 0.74) and RMSE (0.08 versus 0.07), with multi-energy performing slightly better. The images had unique qualitative features: multi-energy consistently showed low-frequency mottling and remnant edge enhancement, whereas multi-distance had more uniform noise texture but variable accuracy and edge blurring. Multi-energy succeeded for all inputs, while multi-distance failed 42.9% of cases.
Conclusion: Material decomposition is traditionally performed with multi-energy input; however, with phase contrast, multi-distance input is possible. At the cost of algorithmic stability, multi-distance decomposition outperforms multi-energy at low dose. The possibility of a stable multi-distance method should be explored to realize the potential advantages of this technique.