Comparison of different numerical treatments for x-ray phase tomography of soft tissue from differential phase projections

Abstract

X-ray imaging of soft tissue is made difficult by their low absorbance. The use of x-ray phase imaging and tomography can significantly enhance the detection of these tissues and several approaches have been proposed to this end. Methods such as analyzer-based imaging or grating interferometry produce differential phase projections that can be used to reconstruct the 3D distribution of the sample refractive index. We report on the quantitative comparison of three different methods to obtain x-ray phase tomography with filtered back-projection from differential phase projections in the presence of noise. The three procedures represent different numerical approaches to solve the same mathematical problem, namely phase retrieval and filtered back-projection. It is found that obtaining individual phase projections and subsequently applying a conventional filtered back-projection algorithm produces the best results for noisy experimental data, when compared with other procedures based on the Hilbert transform. The algorithms are tested on simulated phantom data with added noise and the predictions are confirmed by experimental data acquired using a grating interferometer. The experiment is performed on unstained adult zebrafish, an important model organism for biomedical studies. The method optimization described here allows resolution of weak soft tissue features, such as muscle fibers.

X-ray Multimodal Tomography Using Speckle-Vector Tracking

“…The phase-imaging method employed here is equivalent to x-ray grating interferometry or analyzer-based imaging as the measured signal is the differential phase signal ∇φ [6,8,28]. In these methods, a specific filter is often used in the inverse Radon transform to account for the specificity of the signal and perform a one-dimensional integration in the Fourier space [29,30]. Here, as the XSVT technique provides the two transverse differential phase maps, in contrast to the previous techniques, the phase images can be recovered by 2D integration by matrix inversion via, for instance, the Cholesky decomposition.…”

Optimization of contrast and dose in x-ray phase-contrast tomography with a Talbot-Lau interferometer

“…Since only the first derivatives of the total phase shifts in the object are measured with a TLI, an additional integration step is needed before the reconstruction of these differential phase-contrast (DPC) projections. We used the method described in (Pelliccia et al 2015), where integration is achieved by applying a specific Fourier space filter to the DPC projections.…”