Performance characterization of a prototype dual-layer cone-beam computed tomography system.

PURPOSE: Conventional cone-beam computed tomography CT (CBCT) provides limited discrimination between low-contrast tissues. Furthermore, it is limited to full-spectrum energy integration. A dual-energy CBCT system could be used to separate photon energy spectra with the potential to increase the visibility of clinically relevant features and acquire additional information relevant in a multitude of clinical imaging applications. In this work, the performance of a novel dual-layer dual-energy CBCT (DL-DE-CBCT) C-arm system is characterized for the first time. METHODS: A prototype dual-layer detector was fitted into a commercial interventional C-arm CBCT system to enable DL-DE-CBCT acquisitions. DL-DE reconstructions were derived from material-decomposed Compton scatter and photoelectric base functions. The modulation transfer function (MTF) of the prototype DL-DE-CBCT was compared to that of a commercial CBCT. Noise and uniformity characteristics were evaluated using a cylindrical water phantom. Effective atomic numbers and electron densities were estimated in clinically relevant tissue substitutes. Iodine quantification was performed (for 0.5-15 mg/ml concentrations) and virtual noncontrast (VNC) images were evaluated. Finally, contrast-to-noise ratios (CNR) and CT number accuracies were estimated. RESULTS: The prototype and commercial CBCT showed similar spatial resolution, with a mean 10% MTF of 5.98 cycles/cm and 6.28 cycles/cm, respectively, using a commercial standard reconstruction. The lowest noise was seen in the 80 keV virtual monoenergetic images (VMI) (7.40 HU) and the most uniform images were seen at VMI 60 keV (4.74 HU) or VMI 80 keV (1.98 HU), depending on the uniformity measure used. For all the tissue substitutes measured, the mean accuracy in effective atomic number was 98.2% (SD 1.2%) and the mean accuracy in electron density was 100.3% (SD 0.9%). Iodine quantification images showed a mean difference of -0.1 (SD 0.5) mg/ml compared to the true iodine concentration for all blood and iodine-containing objects. For VNC images, all blood substitutes containing iodine averaged a CT number of 43.2 HU, whereas a blood-only substitute measured 44.8 HU. All water-containing iodine substitutes measured a mean CT number of 2.6 in the VNC images. A noise-suppressed dataset showed a CNR peak at VMI 40 keV and low at VMI 120 keV. In the same dataset without noise suppression applied, a peak in CNR was obtained at VMI 70 keV and a low at VMI 120 keV. The estimated CT numbers of various clinically relevant objects were generally very close to the calculated CT number. CONCLUSIONS: The performance of a prototype dual-layer dual-energy C-arm CBCT system was characterized. Spatial resolution and noise were comparable with a commercially available C-arm CBCT system, while offering dual-energy capability. Iodine quantifications, effective atomic numbers, and electron densities were in good agreement with expected values, indicating that the system can be used to reliably evaluate the material composition of clinically relevant tissues. The VNC and monoenergetic images indicate a consistent ability to separate clinically relevant tissues. The results presented indicate that the system could find utility in diagnostic, interventional, and radiotherapy planning settings.

ITERATIVE SCATTER CORRECTION FOR GRID-LESS BEDSIDE CHEST RADIOGRAPHY: PERFORMANCE FOR A CHEST PHANTOM.

The aim of this work was to experimentally compare the contrast improvement factors (CIFs) of a newly developed software-based scatter correction to the CIFs achieved by an antiscatter grid. To this end, three aluminium discs were placed in the lung, the retrocardial and the abdominal areas of a thorax phantom, and digital radiographs of the phantom were acquired both with and without a stationary grid. The contrast generated by the discs was measured in both images, and the CIFs achieved by grid usage were determined for each disc. Additionally, the non-grid images were processed with a scatter correction software. The contrasts generated by the discs were determined in the scatter-corrected images, and the corresponding CIFs were calculated. The CIFs obtained with the grid and with the software were in good agreement. In conclusion, the experiment demonstrates quantitatively that software-based scatter correction allows restoring the image contrast of a non-grid image in a manner comparable with an antiscatter grid.

A MONTE-CARLO SIMULATION FRAMEWORK FOR JOINT OPTIMISATION OF IMAGE QUALITY AND PATIENT DOSE IN DIGITAL PAEDIATRIC RADIOGRAPHY.

In paediatric radiography, according to the as low as reasonably achievable (ALARA) principle, the imaging task should be performed with the lowest possible radiation dose. This paper describes a Monte-Carlo simulation framework for dose optimisation of imaging parameters in digital paediatric radiography. Patient models with high spatial resolution and organ segmentation enable the simultaneous evaluation of image quality and patient dose on the same simulated radiographic examination. The accuracy of the image simulation is analysed by comparing simulated and acquired images of technical phantoms. As a first application example, the framework is applied to optimise tube voltage and pre-filtration in newborn chest radiography. At equal patient dose, the highest CNR is obtained with low-kV settings in combination with copper filtration.

X-ray imaging performance of scintillator-filled silicon pore arrays.

The need for fine detail visibility in various applications such as dental imaging, mammography, but also neurology and cardiology, is the driver for intensive efforts in the development of new x-ray detectors. The spatial resolution of current scintillator layers is limited by optical diffusion. This limitation can be overcome by a pixelation, which prevents optical photons from crossing the interface between two neighboring pixels. In this work, an array of pores was etched in a silicon wafer with a pixel pitch of 50 microm. A very high aspect ratio was achieved with wall thicknesses of 4-7 microm and pore depths of about 400 microm. Subsequently, the pores were filled with Tl-doped cesium iodide (CsI:Tl) as a scintillator in a special process, which includes powder melting and solidification of the CsI. From the sample geometry and x-ray absorption measurement the pore fill grade was determined to be 75%. The scintillator-filled samples have a circular active area of 16 mm diameter. They are coupled with an optical sensor binned to the same pixel pitch in order to measure the x-ray imaging performance. The x-ray sensitivity, i.e., the light output per absorbed x-ray dose, is found to be only 2.5%-4.5% of a commercial CsI-layer of similar thickness, thus very low. The efficiency of the pores to transport the generated light to the photodiode is estimated to be in the best case 6.5%. The modulation transfer function is 40% at 4 lp/mm and 10%-20% at 8 lp/mm. It is limited most likely by the optical gap between scintillator and sensor and by K-escape quanta. The detective quantum efficiency (DQE) is determined at different beam qualities and dose settings. The maximum DQE(0) is 0.28, while the x-ray absorption with the given thickness and fill factor is 0.57. High Swank noise is suspected to be the reason, mainly caused by optical scatter inside the CsI-filled pores. The results are compared to Monte Carlo simulations of the photon transport inside the pore array structure. In addition, some x-ray images of technical and anatomical phantoms are shown. This work shows that scintillator-filled pore arrays can provide x-ray imaging with high spatial resolution, but are not suitable in their current state for most of the applications in medical imaging, where increasing the x-ray doses cannot be tolerated.