1156 - Metal Identification and Artifact Reduction Using a Dual-Layer CBCT Imager Prototype

M. Jacobson¹, T. C. Harris², D. Ferguson³, M. Lehmann⁴, R. Bruegger⁴, J. OConnell¹, M. Myronakis³, P. Corral Arroyo⁴, Y. H. Hu⁵, V. Birrer⁴, R. Fueglistaller⁴, D. Morf⁶, and R. I. Berbeco¹; ¹Department of Radiation Oncology, Brigham and Women’s Hospital, Dana Farber Cancer Institute and Harvard Medical School, Boston, MA, ²Department of Physics, University of Heidelberg, Heidelberg, Germany, ³Brigham and Womens Hospital, Boston, MA, ⁴Varian Medical Systems, Baden-Dattwil, Switzerland, ⁵Department of Radiation Oncology, Brigham and Women’s Hospital/Dana-Farber Cancer Institute, Boston, MA, ⁶Varian Medical Systems, Palo Alto, CA

Purpose/Objective(s): Cone beam CT (CBCT) is common in radiation therapy image guidance. However, when metal prosthetics or implanted fiducials are present in the patient, CBCTs exhibit image artifacts which may impede targeting and registration with prior planning CTs. Likewise, metal artifacts hinder dose recalculation in adaptive radiotherapy (ART). Many artifact suppression algorithms rely on an initial delineation of the metal structures, either in the acquired 2D x-ray projection images or in their 3D reconstruction. Delineation in 2D presents challenges, since overlapping structures in the projections impede recognition of metal. Delineation in 3D is also challenging, since metal artifacts in initially uncorrected CBCT images obscure the boundaries of metal objects. In this work, we test an artifact reduction method employing novel dual-layer imager (DLI) data to facilitate 3D metal delineation. Materials/

Methods: The DLI is composed of two detector layers, each containing CsI scintillator sub-layers of different thicknesses, and an aSi photodiode sub-layer. It was mounted on a clinical linac and used to scan several patients, including a head and neck patient with dental implants and a lower abdominal patient with prostate fiducials. The x-ray projections acquired in each layer were reconstructed without spectral correction and subtracted to form a 3D subtraction image. Metal regions in the subtraction images were both prominent in intensity and sharply delineated, and hence easily thresholded to produce metal segmentation maps. Metal-affected areas of the projections were located through reprojection of the maps and overpainted using a Dirchlet boundary problem solver. Over-painted projections from both layers were averaged together and reconstructed to form the final CBCT images. Voxel standard deviations in regions of interests (ROIs) were used to quantify tissue uniformity in metal-affected locations and to quantify noise in other locations.
Results: In addition to a substantial visual reduction in artifacts, ROI measurements showed significant restoration of tissue uniformity vis-a-vis standard CBCT. Uniformity improved by 30% near the prostate fiducials and by a factor of 9 near the dental implants. Moreover, image noise was 30% lower near the center of the reconstructed lower abdomen, as compared to standard imaging.
Conclusion: These preliminary patient studies found the multi-spectral, geometrically aligned x-ray measurements provided by dual-layer CBCT to be an effective tool for locating and correcting for implanted metal structures. Moreover, the increased radiation capture efficiency of the DLI reduced image noise in large, lower abdominal anatomy. This stands to benefit registration accuracy with pre-treatment imaging as well as on-treatment dose recalculations in ART.