The application of metal artifact reduction (MAR) in CT scans for radiation oncology by monoenergetic extrapolation with a DECT scanner

Abstract

Metal artifacts in computed tomography CT images are one of the main problems in radiation oncology as they introduce uncertainties to target and organ at risk delineation as well as dose calculation. This study is devoted to metal artifact reduction (MAR) based on the monoenergetic extrapolation of a dual energy CT (DECT) dataset. In a phantom study the CT artifacts caused by metals with different densities: aluminum (ρAl=2.7g/cm3), titanium (ρTi=4.5g/cm3), steel (ρsteel=7.9g/cm3) and tungsten (ρW=19.3g/cm3) have been investigated. Data were collected using a clinical dual source dual energy CT (DECT) scanner (Siemens Sector Healthcare, Forchheim, Germany) with tube voltages of 100kV and 140kV(Sn). For each tube voltage the data set in a given volume was reconstructed. Based on these two data sets a voxel by voxel linear combination was performed to obtain the monoenergetic data sets. The results were evaluated regarding the optical properties of the images as well as the CT values (HU) and the dosimetric consequences in computed treatment plans. A data set without metal substitute served as the reference. Also, a head and neck patient with dental fillings (amalgam ρ=10g/cm3) was scanned with a single energy CT (SECT) protocol and a DECT protocol. The monoenergetic extrapolation was performed as described above and evaluated in the same way. Visual assessment of all data shows minor reductions of artifacts in the images with aluminum and titanium at a monoenergy of 105keV. As expected, the higher the densities the more distinctive are the artifacts. For metals with higher densities such as steel or tungsten, no artifact reduction has been achieved. Likewise in the CT values, no improvement by use of the monoenergetic extrapolation can be detected. The dose was evaluated at a point 7cm behind the isocenter of a static field. Small improvements (around 1%) can be seen with 105keV. However, the dose uncertainty remains of the order of 10% to 20%. Thus, the improvement is not significant for radiotherapy planning. For amalgam with a density between steel and tungsten, monoenergetic data sets of a patient do not show substantial artifact reduction. The local dose uncertainties around the metal artifact determined for a static field are of the order of 5%. Although dental fillings are smaller than the phantom inserts, metal artifacts could not be reduced effectively. In conclusion, the image based monoenergetic extrapolation method does not provide efficient reduction of the consequences of CT-generated metal artifacts for radiation therapy planning, but the suitability of other MAR methods will be subsequently studied.