Reduction of motion artefacts in non-gated dual-energy radiography.

The objective of this work was to reduce motion artefacts in non-gated dual-energy subtraction radiography whilst preserving the contrast-to-noise ratio (CNR) in regions with low motion. Dual-energy radiography provides material-selective information (soft-tissue and bone images) that may be used for improved detection of calcifications in lung nodules. The weighted logarithmic dual-energy subtraction of thoracic images performed without electrocardiogram gating results in motion-induced artefacts. The low-energy image was acquired at the usual dose setting at 60 kV. To obtain the high-energy (120 kV) information, a series of consecutive images at a time interval of 30 ms were made. The series integral dose was equivalent to the dose of a single conventional high-energy image. A motion-free merging technique was introduced that combines standard images yielding low image noise with phase-selective images yielding motion artefact-free image regions which are used for dual-energy subtraction. Evaluations of the method were performed with simulations and measurements using a C-arm system (Axiom Artis; Siemens AG, Germany) equipped with a flat detector of 40 x 30 cm(2). The merging approach conserved standard image noise levels and the CNR in areas without cardiac motion, whereas image noise in pericardial lung regions and in the heart was increased compared with standard images. Motion artefacts in the heart and in the lung areas close to the heart are significantly reduced in the material-selective images when compared with a standard non-gated subtraction.

Simultaneous misalignment correction for approximate circular cone-beam computed tomography.

Currently, CT scanning is often performed using flat detectors which are mounted on C-arm units or dedicated gantries as in radiation therapy or micro CT. For perspective cone-beam backprojection of the Feldkamp type (FDK) the geometry of an approximately circular scan trajectory has to be available for reconstruction. If the system or the scan geometry is afflicted with geometrical instabilities, referred to as misalignment, a non-perfect approximate circular scan is the case. Reconstructing a misaligned scan without knowledge of the true trajectory results in severe artefacts in the CT images. Unlike current methods which use a pre-scan calibration of the geometry for defined scan protocols and calibration phantoms, we propose a real-time iterative restoration of reconstruction geometry by means of entropy minimization. Entropy minimization is performed combining a simplex algorithm for multi-parameter optimization and iterative graphics card (GPU)-based FDK-reconstructions. Images reconstructed with the misaligned geometry were used as an input for the entropy minimization algorithm. A simplex algorithm changes the geometrical parameters of the source and detector with respect to the reduction of entropy. In order to reduce the size of the high-dimensional space required for minimization, the trajectory was described by only eight fix points. A virtual trajectory is generated for each iteration using a least-mean-squares algorithm to calculate an approximately circular path including these points. Entropy was minimal for the ideal dataset, whereas strong misalignment resulted in a higher entropy value. For the datasets used in this study, the simplex algorithm required 64-200 iterations to achieve an entropy value equivalent to the ideal dataset, depending on the grade of misalignment using random initialization conditions. The use of the GPU reduced the time per iteration as compared to a quad core CPU-based backprojection by a factor of 10 resulting in a total of 15-20 ms per iteration, and thus providing an online geometry restoration after a total computation time of approximately 1-3 s, depending on the number of iterations. The proposed method provides accurate geometry restoration for approximately circular scans and eliminates the need for an elaborate off-line calibration for each scan. If a priori information about the trajectory is used to initialize the simplex algorithm, it is expected that the entropy minimization will converge significantly faster.