PT - JOURNAL ARTICLE AU - Gaspar Delso AU - Rebekka Kraus AU - Ralph Bundschuh AU - Axel Martinez AU - Stephan Nekolla AU - Sibylle Ziegler TI - PET/MR truncation artifact reduction with an active model of the patient’s arms DP - 2010 May 01 TA - Journal of Nuclear Medicine PG - 1379--1379 VI - 51 IP - supplement 2 4099 - http://jnm.snmjournals.org/content/51/supplement_2/1379.short 4100 - http://jnm.snmjournals.org/content/51/supplement_2/1379.full SO - J Nucl Med2010 May 01; 51 AB - 1379 Objectives Simultaneous PET/MR has great potential for clinical diagnosis and research. However, MR images often do not contain the entire patient’s span, causing truncation artifacts in the PET. Obtaining the patient’s profile from the PET reduces these artifacts, but introduces severe local errors if anatomical constraints are not imposed on the segmentation. A model-based algorithm was tested for this purpose. Methods Ten PET/CT datasets of oncology patients with arms down were used in this study. The attenuation maps were reduced to the expected FOV of an integrated PET/MR (Ø400 - 450 mm) The model was defined by a reduced list of control nodes (gen. < 10) joined by spline interpolation. It was initialized by contour estimation based on max. projections. The evolution of the model was based on a greedy approach cycling through a set of local search operations. The model was first used to locate the patient torso, then the best fit to the partial arms data was computed. Size and position restrictions were imposed based on the segmented torso. The reconstruction was tested using different tissue configurations to fill the model. The results were compared with those of a geometrical model and those of an unconstrained segmentation. Results The model was capable of automatically fitting the arms of all patients. Truncation artifacts were reduced to clinically acceptable values (< 10% error wrt. original) Artifact reduction was equivalent to both unconstrained methods, while strong local errors were eliminated in all but the elbow slices. The estimation of the arm’s profile was shown to have a higher impact than the estimation of the missing tissue type. One cubic centimeter of misplaced tissue led to relative errors of 1.5%, whereas assuming a uniform attenuation of 0.1cm-1 only led to 2.5% errors for a 400mm FOV. Conclusions These results prove that model-based extraction of the patient profile can be effectively used for the correction of PET/MR truncation artifacts. Further work is required for the model to adapt to the patient’s elbow position