The accuracy of ED measurement, Z determination, and iodine quantitation derived from DL-DECT was demonstrated with phantom measurements. The accuracies were not sensitive to scan and reconstruction parameters, namely tube potential, dose, rotation time, and spectral reconstruction level, especially in the case of electron density.

Purpose: Severe artifacts in kilovoltage-CT simulation images caused by large metallic implants can significantly degrade the conspicuity and apparent CT Hounsfield number of targets and anatomic structures, jeopardize the confidence of anatomical segmentation, and introduce inaccuracies into the radiation therapy treatment planning process. This study evaluated the performance of the first commercial orthopedic metal artifact reduction function (O-MAR) for radiation therapy, and investigated its clinical applications in treatment planning. Methods: Both phantom and clinical data were used for the evaluation. The CIRS electron density phantom with known physical (and electron) density plugs and removable titanium implants was scanned on a Philips Brilliance Big Bore 16-slice CT simulator. The CT Hounsfield numbers of density plugs on both uncorrected and O-MAR corrected images were compared. Treatment planning accuracy was evaluated by comparing simulated dose distributions computed using the true density images, uncorrected images, and O-MAR corrected images. Ten CT image sets of patients with large hip implants were processed with the O-MAR function and evaluated by two radiation oncologists using a five-point score for overall image quality, anatomical conspicuity, and CT Hounsfield number accuracy. By utilizing the same structure contours delineated from the O-MAR corrected images, clinical IMRT treatment plans for five patients were computed on the uncorrected and O-MAR corrected images, respectively, and compared. Results: Results of the phantom study indicated that CT Hounsfield number accuracy and noise were improved on the O-MAR corrected images, especially for images with bilateral metal implants. The γ pass rates of the simulated dose distributions computed on the uncorrected and O-MAR corrected images referenced to those of the true densities were higher than 99.9% (even when using 1% and 3 mm distance-to-agreement criterion), suggesting that dose distributions were clinically identical. In all patient cases, radiation oncologists rated O-MAR corrected images as higher quality. Formerly obscured critical structures were able to be visualized. The overall image quality and the conspicuity in critical organs were significantly improved compared with the uncorrected images: overall quality score (1.35 vs 3.25, P = 0.0022); bladder (2.15 vs 3.7, P = 0.0023); prostate and seminal vesicles/vagina (1.3 vs 3.275, P = 0.0020); rectum (2.8 vs 3.9, P = 0.0021). The noise levels of the selected ROIs were reduced from 93.7 to 38.2 HU. On most cases (8/10), the average CT Hounsfield numbers of the prostate/vagina on the O-MAR corrected images were closer to the referenced value (41.2 HU, an average measured from patients without metal implants) than those on the uncorrected images. High γ pass rates of the five IMRT dose distribution pairs indicated that the dose distributions were not significantly affected by the CT image improvements. Conclusions: Overall, this study indicated that the O-MAR function can remark...

Respiration can cause tumors in the thorax or abdomen to move by as much as 3 cm; this movement can adversely affect the planning and delivery of radiation treatment. Several techniques have been used to compensate for respiratory motion, but all have shortcomings. Manufacturers of computed tomography (CT) equipment have recently used a technique developed for cardiac CT imaging to track respiratory-induced anatomical motion and to sort images according to the phase of the respiratory cycle they represent. Here we propose a method of generating CT images that accounts for respiratory-induced anatomical motion on the basis of displacement, i.e., displacement-binned CT image sets. This technique has shown great promise, however, it is not fully supported by currently used CT image reconstruction software. As an interim solution, we have developed a method for extracting displacement-binned CT image data sets from data sets assembled on the basis of a prospectively determined breathing phase acquired on a multislice helical CT scanner. First, the projection data set acquired from the CT scanner was binned at small phase intervals before reconstruction. The manufacturer's software then generated image sets identified as belonging to particular phases of the respiratory cycle. All images were then individually correlated to the displacement of an external fiducial marker. Next, CT image data sets were resorted on the basis of the displacement and assigned an appropriate phase. Finally, displacement-binned image data sets were transferred to a treatment-planning system for analysis. Although the technique is currently limited by the phase intervals allowed by the CT software, some improvement in image reconstruction was seen, indicating that this technique is useful at least as an interim measure.