Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf

2 Publications

Methods for an Automatic Analysis of Motion Tracking Data in PET

Langner, J.; Oehme, L.; Pötzsch, C.; Dittrich, S.; Beuthien-Baumann, B.; van den Hoff, J.

Due to the constantly improving spatial resolution of PET systems, patient motion increasingly limits the achievable image quality in PET. Especially for tracer kinetic analysis in dynamic PET, as well as for the analysis of small spatial regions of interests (ROI), patient motion represents a severe obstacle. Therefore, different methods for the tracking and correction of patient movement have been investigated in the past. Generally, suitable motion threshold values have to be deduced from the motion data to identify significant motion and to reduce the amount of computation time for motion compensation methods. Therefore, an objective analysis of the motion data has to be performed. Motion data is usually provided in the form of three translations along, and three rotations around, the coordinate axes. These raw parameters, however, proof to be unsuited for a direct assessment of the magnitude of 3D motion. Rather, it is necessary to execute the spatial transformations defined by the six parameters for each point within the ROI. Therefore, we developed procedures for an automated analysis of motion tracking data which address these problems. The raw motion data from the tracking device is automatically processed via methods based on the 'R' statistics toolkit. Parameters are calculated for the general stability of the motion tracking. Furthermore, the translation and rotation parameters are analysed and graphically displayed. Time dependent transformation matrices relative to the acquisition start are also calculated. For the quantitative analysis of head motion, we translate the motion information onto the surface of a virtual sphere with a diameter comparable to that of a human brain (~20 cm). By transforming each point of a grid on the sphere with the available motion parameters, the 3D-Euclidean distance of each grid point from its original position is calculated. If the distance of any of these points exceeds a certain threshold (~3 mm), the motion is considered to be significant.
Furthermore, a target ROI (e.g. the striatum) can be specified. This analysis allows the objective identification of all time points where significant motion occurred. This enables optimised settings for motion correction algorithms, notably avoiding time intensive redundant computations.
Furthermore, this data enables an intuitive quality control by providing pseudo 3D plots of the motion on the reference sphere. Finally, a report page is provided yielding summary information concerning the areas of largest/smallest motion as well as an automatic motion score of the study as a feedback for the physicians.

  • Poster
    EANM'08, Annual Congress of the European Association of Nuclear Medicine, 11.-15.10.2008, München, Deutschland
  • Abstract in refereed journal
    European Journal of Nuclear Medicine and Molecular Imaging 35(2008), S333
    DOI: 10.1007/s00259-008-0896-9

Publ.-Id: 11394