Computed tomography (CT) reconstruction from X-ray projections acquired within a limited angle range is challenging. Both analytical and conventional iterative methods suffer from severe artifacts due to the incompleteness of sinogram. To obtain high-quality reconstructions from limited angle CT, it is crucial to integrate model-based methods with better learned priors from existing big databases of CT images. Transform learning is an unsupervised data-driven model that has recently shown promise in several medical imaging applications. However, its performance is limited due to the use of hand-crafted penalty terms on the learned transform and sparse coefficients. Inspired by the great success of convolutional neural network, we propose a supervised transform learning method for limited angle CT image reconstruction, where we redesign the conventional unsupervised iterative transform learning algorithm and learn the priors for both sparse coefficients and transform in a supervised manner. Clinical patient data results show that the proposed method significantly improves image quality of reconstructions, compared to a denoising deep convolutional neural network method, FBPConvNet, and a representative iterative neural network method, LEARN.
Quantitative photoacoustic tomography (QPAT) is a hybrid imaging modality that simultaneously reconstructs absorption and scattering coefficients with multi-source or multi-wavelength setting. In contrast with PAT, QPAT eliminates the artifacts due to photon transfer in depth and characterizes the intrinsic biological attributes. However, data acquisition for QPAT is time consuming because for each optical source (or wavelength), a 360-degree acoustic measurement on the boundary of imaging region is required. In this work, we investigate a novel limited-view multi-source (LVMS) QPAT scheme and reconstruction algorithm based on coupled opto-acousto model. A unique setting that binds optical source and acoustic detector together is presented. Under each illumination, only a limited-view measurement is acquired, which is incomplete for PAT reconstruction but sufficient for direct QPAT reconstruction; then the source and detector rotate to next data acquisition position synchronously. In terms of reconstruction, a sparsity-regularized formulation based on total variation norm is adopted here and optimized through alternating direction method of multipliers with LBFGS solver, during which the adjoint method is used for rapid computation of numerical gradient of objective function. However, the aperture effect i.e. anisotropic angular sensitivity of a finite-dimension transducer would cause temporal distortion of receiving acoustic signal and further resolution reduction. Therefore, we numerically integrate ideal PAT system spatial impulse response (SIR) with the angular response from circular transducer to improve the modeling accuracy. In summary, the proposed LVMS-QPAT has the potential to shorten the data acquisition time, enhance system SNR and improve image resolution.
We introduce an entirely new technique, termed Photo-Magnetic Imaging (PMI), which overcomes the limitation of pure optical imaging and provides optical absorption at MRI spatial resolution. PMI uses laser light to heat the medium under investigation and employs MR thermometry for the determination of spatially resolved optical absorption in the probed medium. A FEM-based PMI forward solver has been developed by modeling photon migration and heat diffusion in tissue to compare simulation results with measured MRI maps. We have successfully performed PMI using 2.5 cm diameter agar phantom with two low optical absorption contrast (x 4) inclusions under the ANSI limit. Currently, we are developing the PMI inverse solver and undertaking further phantom and in vivo experiments.
4D spatiotemporal images can be naturally divided into the background component, which is temporally coherent, and the motion component, which is spatially sparse, up to the proper basis. And this divide-and-conquer decomposition is an effective sparse representation of 4D images for the purpose of image reconstruction. Based on this prior fact, we introduce Prior Rank, Intensity and Sparsity Model (PRISM): the temporal coherence of the background component is enforced by the rank regularization and the spatial sparsity of the motion component is promoted by the sparsity regularization. In particular, the framelet based PRISM with the multi-resolution and multi-filtered structure will be utilized for image reconstruction. The superior performance of PRISM will be demonstrated with a few new medical imaging applications, including 4D cone beam CT, spiral MRI, and fused MRI-CT multi-modality.
A graphics processing unit-based parallel multigrid solver for a radiative transfer equation with vacuum boundary condition or reflection boundary condition is presented for heterogeneous media with complex geometry based on two-dimensional triangular meshes or three-dimensional tetrahedral meshes. The computational complexity of this parallel solver is linearly proportional to the degrees of freedom in both angular and spatial variables, while the full multigrid method is utilized to minimize the number of iterations. The overall gain of speed is roughly 30 to 300 fold with respect to our prior multigrid solver, which depends on the underlying regime and the parallelization. The numerical validations are presented with the MATLAB codes at https://sites.google.com/site/rtefastsolver/.
Local fine representation of the fluorescence map on the standard mesh can be redundant in the sense that the
reconstruction resolution is usually limited in such a severely ill-posed problem. Using global characteristic shape
functions that can approximately capture the major structural information, we study fluorescence tomography with a new
shape-guided representation based on some underlying mesh. Moreover, the proposed method can naturally enforce the
prior coexistence of fluorescence yield and lifetime when fluorescence maps are formulated in complex sources. The
simulation results suggest that, compared with standard pixel-wise representation, the shape-guided representation offers
better localization of inclusions with improved quantitative accuracy, particularly in the case with inclusions of low
fluorescence contrast, such as 2:1 inclusion-to-background ratio, and is more robust to the initial guess and the noise.
Based on a multigrid forward solver of radiative transfer equation for optical imaging, an efficient multilevel
simultaneous reconstruction of absorption and scattering coefficient is presented, in which L1 minimization can
be used to localize the unknowns, especially in the presence of sparse unknowns.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.