The patient motion can damage the quality of computed tomography images, which are typically acquired in cone-beam geometry. The rigid patient motion is characterized by six geometric parameters and are more challenging to correct than in fan-beam geometry. We extend our previous rigid patient motion correction method based on the principle of locally linear embedding (LLE) from fan-beam to cone-beam geometry and accelerate the computational procedure with the graphics processing unit (GPU)-based all scale tomographic reconstruction Antwerp toolbox. The major merit of our method is that we need neither fiducial markers nor motion-tracking devices. The numerical and experimental studies show that the LLE-based patient motion correction is capable of calibrating the six parameters of the patient motion simultaneously, reducing patient motion artifacts significantly.
Artifacts resulting from metal objects have been a persistent problem in CT images over the last four decades. A common
approach to overcome their effects is to replace corrupt projection data with values synthesized from an interpolation
scheme or by reprojection of a prior image. State-of-the-art correction methods, such as the interpolation- and
normalization-based algorithm NMAR, often do not produce clinically satisfactory results. Residual image artifacts remain
in challenging cases and even new artifacts can be introduced by the interpolation scheme. Metal artifacts continue to be
a major impediment, particularly in radiation and proton therapy planning as well as orthopedic imaging. A new solution
to the long-standing metal artifact reduction (MAR) problem is deep learning, which has been successfully applied to
medical image processing and analysis tasks. In this study, we combine a convolutional neural network (CNN) with the
state-of-the-art NMAR algorithm to reduce metal streaks in critical image regions. Training data was synthesized from CT
simulation scans of a phantom derived from real patient images. The CNN is able to map metal-corrupted images to
artifact-free monoenergetic images to achieve additional correction on top of NMAR for improved image quality. Our
results indicate that deep learning is a novel tool to address CT reconstruction challenges, and may enable more accurate
tumor volume estimation for radiation therapy planning.
The rapidly-rising field of machine learning, including deep learning, has inspired applications across many disciplines. In medical imaging, deep learning has been primarily used for image processing and analysis. In this paper, we integrate a convolutional neural network (CNN) into the computed tomography (CT) image reconstruction process. Our first task is to monitor the quality of CT images during iterative reconstruction and decide when to stop the process according to an intelligent numerical observer instead of using a traditional stopping rule, such as a fixed error threshold or a maximum number of iterations. After training on ground truth images, the CNN was successful in guiding an iterative reconstruction process to yield high-quality images. Our second task is to improve a sinogram to correct for artifacts caused by metal objects. A large number of interpolation and normalization-based schemes were introduced for metal artifact reduction (MAR) over the past four decades. The NMAR algorithm is considered a state-of-the-art method, although residual errors often remain in the reconstructed images, especially in cases of multiple metal objects. Here we merge NMAR with deep learning in the projection domain to achieve additional correction in critical image regions. Our results indicate that deep learning can be a viable tool to address CT reconstruction challenges.
For early detection and targeted therapy, receptor expression profiling is instrumental to classifying breast cancer into
sub-groups. In particular, human epidermal growth factor receptor 2 (HER2) expression has been shown to have both
prognostic and predictive values. Recently, an increasingly more complex view of HER2 in breast cancer has emerged
from genome sequencing that highlights the role of inter- and intra-tumor heterogeneity in therapy resistance. Studies on
such heterogeneity demand high-content, high-resolution functional and molecular imaging in vivo, which cannot be
achieved using any single imaging tool. Clearly, there is a critical need to develop a multimodality approach for breast
cancer imaging. Since 2006, grating-based x-ray imaging has been developed for much-improved x-ray images. In 2014,
the demonstration of fluorescence molecular tomography (FMT) guided by x-ray grating-based micro-CT was reported
with encouraging results and major drawbacks. In this paper, we propose to integrate grating-based x-ray tomography
(GXT) and high-dimensional optical tomography (HOT) into the first-of-its-kind truly-fused GXT-HOT (pronounced as
“Get Hot”) system for imaging of breast tumor heterogeneity, HER2 expression and dimerization, and therapeutic
response. The primary innovation lies in developing a brand-new high-content, high-throughput x-ray optical imager
based on several contemporary techniques to have MRI-type soft tissue contrast, PET-like sensitivity and specificity, and
micro-CT-equivalent resolution. This system consists of two orthogonal x-ray Talbot-Lau interferometric imaging chains
and a hyperspectral time-resolved single-pixel optical imager. Both the system design and pilot results will be reported in
this paper, along with relevant issues under further investigation.
X-ray phase contrast imaging is an important mode due to its sensitivity to subtle features of soft biological tissues.
Grating-based differential phase contrast (DPC) imaging is one of the most promising phase imaging techniques because
it works with a normal x-ray tube of a large focal spot at a high flux rate. However, a main obstacle before this paradigm
shift is the fabrication of large-area gratings of a small period and a high aspect ratio. Imaging large objects with a size-limited
grating results in data truncation which is a new type of the interior problem. While the interior problem was
solved for conventional x-ray CT through analytic extension, compressed sensing and iterative reconstruction, the
difficulty for interior reconstruction from DPC data lies in that the implementation of the system matrix requires the
differential operation on the detector array, which is often inaccurate and unstable in the case of noisy data. Here, we
propose an iterative method based on spline functions. The differential data are first back-projected to the image space.
Then, a system matrix is calculated whose components are the Hilbert transforms of the spline bases. The system matrix
takes the whole image as an input and outputs the back-projected interior data. Prior information normally assumed for
compressed sensing is enforced to iteratively solve this inverse problem. Our results demonstrate that the proposed
algorithm can successfully reconstruct an interior region of interest (ROI) from the differential phase data through the
X-ray tensor tomography is a promising imaging modality for probing the micro structure of a sample by reconstructing
small-angle scattering densities in different scattering directions. However, the current x-ray grating technique still faces
an obstacle when a divergent x-ray beam from a point x-ray source propagates through a large object and reaches large
planar gratings. In this situation, tensor interior tomography is essential to perform the image reconstruction over a
region of interest (ROI) in the object. Therefore, we propose interior tensor tomography with 2D gratings to extract dark
field images isotropically. Our numerical results demonstrate that the proposed methods are promising for reconstruction
of local images from truncated dark field projection data.
Recent advances in X-ray imaging technologies have paved the way for use of energy-discriminating photon-counting
detector arrays. These detectors show promise in clinical and preclinical applications. Multi-energy or spectral CT images
can be visualized in multi-colors. Despite the advantages offered by the spectral dimension of acquired data, higher image
resolution is still desirable, especially in challenging tasks such as on-site studies of resected pathological tissues. Here we
propose to enhance image resolution of a spectral X-ray imaging system by partially blocking each detector element with
an absorption grating (for reduced aperture), commonly used for Talbot-Lau interferometry. After acquiring X-ray data at
an initial grating-detector configuration, the grating is shifted to expose previously blocked portions so that each
measurement contains new information. All the acquired data are then combined into an augmented system matrix and
subsequently reconstructed using an iterative algorithm. Our proof of concept simulations are performed with MCNP6.1
code and the experiment was performed using a Hamamatsu microfocus X-ray source, an absorption grating, and an Xray
camera. Our results demonstrate that the gratings commonly used for x-ray phase-contrast imaging have a utility for
super-resolution imaging performance.