[1] Wenxiang Cong, et al, “Tomographic Image Reconstruction Using an Advanced Score Function (ADSF),” arXiv:2306.08610(2023)
Leveraging the Gaussian mixture model, we derive a novel score matching formula to establish an advanced score function (ADSF) through deep learning. Using the ADSF, a new iterative reconstruction method is developed to improve image reconstruction. The performance of the reconstruction method is evaluated on clinical raw CT datasets, showing excellent generalizability and stability.
[2] Wenxiang Cong, et al, “Virtual monoenergetic CT imaging via deep learning,” Patterns 1(8), 100128(2020).
We have developed a deep learning approach to map polyenergetic CT images to monoenergetic images at pre-specified energy levels. To efficiently perform this task, we propose an improved residual neural network (IResNet) model. Trained on clinical dual-energy CT (DECT) data, the IResNet demonstrates excellent convergence behavior and produces accurate virtual monoenergetic (VM) images comparable to those generated by DECT. This enables multi-material decomposition with performance on par with DECT.
[3] Wenxiang Cong, et al, “CT image reconstruction on a low dimensional manifold,” Inverse Problems and Imaging 13(3), 449-460 (2019).
We apply the low-dimensional manifold model (LDMM) to regularize X-ray computed tomography (CT) image reconstruction. This approach effectively recovers detailed structural information, significantly enhancing both spatial and contrast resolution. Clinical experiments show that the LDMM-based method achieves more accurate image reconstruction with high fidelity and improved contrast resolution.
[4] Wenxiang Cong, et al, “Spherical grating based x‐ray Talbot interferometry,” Medical physics 42(11), 6514-6519 (2015).
A spherical grating matches the wave front of a point x-ray source very well, allowing the perpendicular incidence of x-rays on the grating to achieve a higher visibility over a larger field of view than the planer grating counterpart. A theoretical analysis of the Talbot effect for spherical grating imaging is proposed to establish a basic foundation for x-ray spherical gratings interferometry. An efficient method of spherical grating imaging is also presented to extract attenuation, differential phase, and dark-field images in the x-ray spherical grating interferometer.
[5] Wenxiang Cong, et al, “Differential phase-contrast interior tomography,” Physics in Medicine & Biology 57(10), 2905-2914 (2012).
Differential phase-contrast interior tomography enables the reconstruction of refractive index distributions within a region of interest (ROI) for the visualization and analysis of structures inside large biological specimens. In this imaging mode, X-ray scanning targets only the ROI, with a narrow beam passing through the object, significantly reducing both radiation dose and system cost. We demonstrate through numerical analysis that accurate interior reconstruction can be achieved from truncated differential projection data of the ROI. This is accomplished via total variation minimization, assuming a piecewise constant refractive index distribution within the ROI. We then develop a practical iterative algorithm for this interior reconstruction and conduct numerical experiments to validate the feasibility of the proposed approach.
[6] Wenxiang Cong, et al, “Modeling photon propagation in biological tissues using a generalized Delta-Eddington phase function,” Physical Review E 76(5), 051913 (2007).
Photon propagation in biological tissue is commonly described by the radiative transfer equation, where the phase function represents the scattering characteristics of the medium and significantly affects both the accuracy of the solution and computational efficiency. In this work, we introduce a generalized Delta-Eddington phase function to simplify the radiative transfer equation into an integral equation for photon fluence rate. Compared to the widely used diffusion approximation model, the solution of this integral equation offers highly accurate modeling of photon propagation in biological tissue across a broad range of optical parameters. This methodology is validated through Monte Carlo simulation.