Monte Carlo simulation of scattering in X-ray phase contrast imaging
Recrutement: 
Recrutement en cours/passé: 
Recrutement passé
Periode: 
2019
Contact: 
Max Langer : max.langer@creatis.insa-lyon.fr Jean Michel Letang : Jean-Michel.Letang@creatis.insa-lyon.fr

Monte Carlo simulation of X-ray phase contrast for mammographic imaging

X-ray phase contrast imaging permits to reach nanometric resolution in tomographic imaging with several orders of magnitude higher sensitivity than using the attenuation. The main drawback is that it needs an additional reconstruction step, known as phase retrieval, to yield quantitative images. Several methods for generating phase contrast have been developed, the most recent one being X-ray speckle-based imaging.

Currently, there are no realistic simulators of X-ray phase contrast. This would have several benefits: optimisation of imaging conditions and reconstruction, planning of experiments, investigation of artefact sources, as well as providing data for machine learning algorithms. Therefore, the aim of this project is to combine simulation of phase contrast with simulation of scattering. Phase contrast is usually modelled from a wave perspective using the Fourier transform, while scattering is usually modelled from a particle perspective using e.g. Monte Carlo simulation.

The main challenge of this project is therefore to combine the two perspectives. In a previous Master project, we developed tools [1] to simulate refraction using MC simulation in Geant4 [2] as well as an analytical phase contrast simulator in GATE [3] and VIP [4].

In a first step, these tools will be applied to speckle-based imaging [5]. Since this is a simpler case, because coherent effects are not necessarily taken into account, the existing code for refraction will be used as a basis for the refinement of the model in the code (voxellised objects, stochastic reflection, real surface...). A numerical emulation of the random mask will have to be defined to simulate speckle imaging. A validation study will be conducted against existing experimental data.

Second, for the diffraction-based applications, the code which currently generates digital reconstructed radiographs (DRR) will have to be combined to scatter. This can be carried out in several ways. The simplest approximation is directly summing the two. A more accurate approach would be first calculating the exit wave-field and phase contrast analytically and calculating the MC probability for position and momentum of the corresponding particles by sampling the Wigner transform of the wave-field.

Besides speckle-based imaging, the developed code will be used to investigate artefact sources in X-ray phase contrast imaging. Phase retrieval from simulated phase contrast images will be used to compare the resulting artefacts with experimental data acquired at the ESRF, with the aim to gain better understanding of artefacts to propose correction algorithms. A common problem with phase retrieval is the presence of low frequency noise in the reconstructions. The sensitivity to noise in the low spatial frequency range is due to low transfer of contrast by the imaging system. The origin of the noise is not known, however. One hypothesis is that it originates from the scattering by the imaged object.

Finally, possibility for use of the code for the simulation of mammographic imaging with the aim of optimising image acquisition protocols and phase reconstruction algorithms will be investigated. This part of the project will be developed in collaboration with Institut Bergonié and AlphaNov, Bordeaux. Future work includes planning of synchrotron experiments to optimise image quality and beam time use, and use of realistic simulated images as training data for machine learning based reconstruction algorithms.

[1] M. Langer et al., Phys. Med. Biol, Submitted

[2] S. Agostinelli et al., Nuclear Instruments and Methods A 506 (2003) 250-303

[3] G Santin et al. IEEE Trans. Nucl. Sci. 50 (2003) 1516-1521

[4] T. Glatard et al., IEEE Transactions on Medical Imaging, 32 (2013) 10-118.

[5] D. Paganin et al., Phys. Rev. A 98, 053813 (2018)